TRIBE

TRIBE - 2025

2025Activity reportProject-TeamTRIBE‌

RNSR: 201923224R

Research center Inria Saclay Centre
Team‌ name: inTeRnet BEyond the usual

Creation of the‌ Project-Team: 2019 June 01

Each year, Inria research‌ teams publish an Activity Report presenting their work‌ and results over the reporting period. These reports‌ follow a common structure, with some optional sections‌ depending on the specific team. They typically begin‌ by outlining the overall objectives and research programme,‌ including the main research themes, goals, and methodological‌ approaches. They also describe the application domains targeted‌ by the team, highlighting the scientific or societal‌ contexts in which their work is situated.

The‌ reports then present the highlights of the year, covering major scientific achievements,‌ software developments, or teaching‌ contributions. When relevant, they‌‌ include sections on software, platforms, and open data,‌ detailing the tools developed‌ and how they are‌‌ shared. A substantial part is dedicated to new‌ results, where scientific contributions‌ are described in detail,‌‌ often with subsections specifying participants and associated keywords.‌

Finally, the Activity Report‌ addresses funding, contracts, partnerships,‌‌ and collaborations at various levels, from industrial agreements‌ to international cooperations. It‌ also covers dissemination and‌‌ teaching activities, such as participation in scientific events,‌ outreach, and supervision. The‌ document concludes with a‌‌ presentation of scientific production, including major publications and‌ those produced during the‌ year.

Keywords

Computer Science‌‌ and Digital Science

A1.1.2. Hardware accelerators (GPGPU, FPGA,‌ etc.)
A1.2.1. Dynamic reconfiguration‌
A1.2.2. Supervision
A1.2.3. Routing‌‌
A1.2.4. QoS, performance evaluation
A1.2.6. Sensor networks
A1.2.8.‌ Network security
A1.2.10. Digital‌ Communications
A1.2.11. Quantum communications‌‌
A1.3.2. Mobile distributed systems
A1.3.3. Blockchain
A1.3.6. Fog,‌ Edge
A1.6. Green Computing‌
A2.3. Embedded and cyber-physical‌‌ systems
A2.3.1. Embedded systems
A2.3.2. Cyber-physical systems
A2.3.5.‌ Cyber-physical systems
A2.5.1. Software‌ Architecture & Design
A2.5.2.‌‌ Component-based Design
A2.5.4. Software Maintenance & Evolution
A2.5.5.‌ Software testing
A2.6.1. Operating‌ systems
A3.1.1. Modeling, representation‌‌
A3.1.3. Distributed data
A3.1.4. Uncertain data
A3.2.2. Knowledge‌ extraction, cleaning
A3.2.3. Inference‌
A3.3.2. Data mining
A4.4.‌‌ Security of equipment and software
A4.8. Privacy-enhancing technologies‌
A5.11.1. Human activity analysis‌ and recognition
A6.1.6. Fractal‌‌ Modeling
A6.2.4. Statistical methods
A6.3.3. Data processing
A7.1.‌ Algorithms
A7.1.3. Graph algorithms‌
A7.1.4. Quantum algorithms
A8.1.‌‌ Discrete mathematics, combinatorics
A8.3. Geometry, Topology
A8.6. Information‌ theory
A8.7. Graph theory‌
A8.9. Performance evaluation
A9.2.‌‌ Machine learning
A9.2.1. Supervised learning
A9.2.2. Unsupervised learning‌
A9.2.3. Reinforcement learning
A9.2.5.‌ Bayesian methods
A9.2.6. Neural‌‌ networks
A9.2.8. Deep learning
A9.6. Decision support
A9.7.‌ AI algorithmics
A9.8. Reasoning‌
A9.9. Distributed AI, Multi-agent‌‌
A9.11. Generative AI

1 Team members, visitors, external collaborators‌

Research Scientists

Aline Carneiro‌ Viana [Team leader‌‌, INRIA, Senior Researcher, HDR]‌
Nadjib Achir [INRIA‌, Researcher, HDR‌‌]
Cédric Adjih [INRIA, Researcher]‌
Emmanuel Baccelli [INRIA‌, Senior Researcher,‌‌ HDR]
Philippe Jacquet [INRIA, Senior‌ Researcher, HDR]‌
Anne Josiane Kouam Djuigne‌‌ [INRIA, Starting Research Position, from‌ May 2025, SRP‌ INRIA + associate researcher‌‌ at TU Berlin]

Post-Doctoral Fellow

Wellington Viana‌ Lobato Junior [INRIA‌, Post-Doctoral Fellow,‌‌ from Feb 2025]

PhD Students

Saeed Alsabbagh‌ [UNIV PARIS SACLAY‌, from May 2025‌‌]
Lucas Airam Castro‌ De Souza [INRIA + UFRJ (Joint PhD)‌]
Lucas Gabriel Da Silva Felix [UFMG‌ (joint PhD with Inria), from Sep 2025‌]
Lucas Gabriel Da Silva Felix [INRIA‌ (joint PhD with UFMG), until Aug 2025‌]
Geoffrey Deperle [INRIA]
Amira Dhaouadi‌ [INRIA]
Felix Marcoccia [CIFRE/Thalès,‌ from Oct 2025, CIFRE]
Rosario Patane‌ [UNIV PARIS SACLAY, until Mar 2025‌]

Technical Staff

Abdelmounaim Bouroudi [INRIA,‌ Engineer, from May 2025]
Mehdi Sofiane‌ Debbah [INRIA, Engineer, until Jun‌ 2025]
Romain Fouquet [INRIA, Engineer‌]
Jeremy Kromer [INRIA, Engineer,‌ from May 2025]
Antoine Lavandier [INRIA‌, Engineer, from May 2025]
Tan‌ Nhat Linh Le [INRIA, Engineer,‌ until Jun 2025]
Fernando Molano Ortiz [‌INRIA, Engineer]
Thanh Son Lam Nguyen‌ [INRIA, Engineer, until Jun 2025‌]
Nils Ponsard [INRIA, Engineer,‌ from May 2025]

Interns and Apprentices

Niruth‌ Savin Bogahawatta [INRIA + USYD, Intern‌, from Nov 2025, 6-month internship]‌
Gustavo Bruno Dos Santos [INRIA, Intern‌, from Dec 2025]
Guillaume Farhi Rivasseau‌ [INRIA, Intern, from Aug 2025‌ until Sep 2025]
Guillaume Farhi Rivasseau [‌INRIA, Intern, until May 2025]‌
Andrei-Valentin Stirbu [INRIA, Intern, from‌ Jul 2025 until Aug 2025]
Marta-Teodora Trales‌ [INRIA, Intern, from Jun 2025‌ until Aug 2025]
Marta-Teodora Trales [INRIA‌, Intern, until Mar 2025]
Mingxuan‌ Wang [INRIA, Intern, from May‌ 2025 until Aug 2025]

Administrative Assistant

Michael‌ Barbosa [INRIA]

Visiting Scientist

Bernard Mans‌ [UNIV MACQUARIE, from Apr 2025 until‌ Jul 2025]

External Collaborators

Guillaume Farhi Rivasseau‌ [ECE PARIS, from Jun 2025 until‌ Jul 2025]
Anne Josiane Kouam Djuigne [‌TU Berlin, until Apr 2025]
Koen‌ Zandberg [Freie Univ Berlin]

2 Overall‌ objectives

TRiBE (“inTeRnet BEyond the Usual”)‌ was officially created in June 2019. TRiBE belongs‌ to the Inria theme “Networks and Telecommunications”.‌ The focus and the evolving directions of TRiBE‌ research contribute, among others, to the priority themes‌ “Digital Security” (as for programming for the Internet‌ of Things) and “Responsible AI and Algorithms” (as‌ for algorithms design, data processing pipelines, green digital‌ twins).

Main goal: Building on a ombination of‌ protocol design, data learning, modeling, and experimental research,‌ TRiBE’s contributions aim to shape smart, unified, and‌ perceptive Internet Edge networks, designed to effectively meet‌ the real demands and purposes of applications, services,‌ and end users, while adapting to the specificities‌ and usability of devices.

2.1 Projections and emerging‌ challenges

The Internet Edge has evolved significantly over‌ the past decades, transitioning from a small, homogeneous network to a vast‌ Internet of Things (IoT)‌ ecosystem, interconnecting a wide‌‌ variety of devices, : while supporting a diverse‌ range of services. We‌ are constantly adapting our‌‌ focus the new related projections:

Significant IoT‌ Devices Growth: The number‌ of global IoT connections‌‌ is expected to rise to nearly 40 billion‌ devices by 2033 (from‌ 16.1 billion in 2023),‌‌ with a compound annual growth rate (CAGR) of‌ 10%. This includes substantial‌ growth in short-range and‌‌ cellular IoT technologies, driven by expanding use‌ cases in consumer (e.g.,‌ smart homes, wearable devices),‌‌ industrial, and urban environments (IoT Analytics;‌ Ericsson IoT Report).‌ Meanwhile, hundreds of billions‌‌ of low-power microcontrollers worldwide are in use daily‌ 70, and tens‌ of billions ship yearly‌‌ 57. These trends bring huge opportunities on‌ the one hand, and‌ on the other hand‌‌ new cybersecurity threats are a major concern 69‌.
Mobile Traffic Growth:‌ Mobile traffic is projected‌‌ to grow exponentially, driven by high-bandwidth applications (e.g.,‌ video streaming, XR) and‌ 5G adoption. Smartphones‌‌ account for 95% of mobile data traffic by‌ 2026, fueled by advancements‌ like video codecs, higher‌‌ screen resolutions, and AI-driven personalized content. Autonomous vehicles‌ are emerging as a‌ significant source of traffic‌‌ demand, relying on HD maps, passenger entertainment systems,‌ and vehicle diagnostics, continuous‌ data exchange with cloud‌‌ servers, HD mapping services, and vehicle-to-everything (V2X) communication‌ (Ericsson Mobility Report‌, Cisco Report,‌‌ V2X White Paper).
5G and Future of‌ 6G: 5G-enabled IoT devices‌ are key to delivering‌‌ high-speed, low-latency applications at the network edge. By‌ 2033, 5.5 billion cellular‌ connections (including mMTC,‌‌ as NB-IoT and LTE-M, and RedCap/eRedCap) are‌ expected, including 1.1 billion‌ full 5G NR connections.‌‌ In Europe, 87% of mobile users will have‌ 5G coverage by 2030.‌ Looking ahead, 6G aims‌‌ to integrate enhanced mobile broadband with AI,‌ driving innovations in industries‌ like smart cities and‌‌ autonomous transportation (N-iX, IEEEComSoc Tech. Blog‌, GSMA Report).‌
Edge Computing and Device‌‌ Specificity: Edge computing, combined with IoT and AI,‌ is revolutionizing many sectors‌ (e.g., smart industries, intelligent‌‌ transportation systems, healthcare, smart cities, industrial automation). By‌ processing data closer to‌ devices, edge computing reduces‌‌ latency and improves efficiency. By 2030, 75% of‌ enterprise-generated data is expected‌ to be created and‌‌ processed outside traditional data centers, highlighting the critical‌ role of Internet edge‌ networks and devices.‌‌ However, while the edge is critical for real-time,‌ localized tasks, the cloud‌ remains essential for large-scale‌‌ data aggregation, advanced AI model training, and global‌ coordination. We are‌ convinced that, together, edge‌‌ and cloud architectures enable a balanced approach, combining‌ real-time decision-making with scalable‌ insights to support the‌‌ growing demands of IoT ecosystems (61,‌ N-iX, IDC Report‌, Transforma Insights).‌‌

These projections emphasize the growing need for smarter,‌ more efficient networking and‌ IoT solutions, along with‌‌ adaptive edge computing, to‌ address the increase in IoT connections, the rising‌ impact of mobile networking connectivity, and the resulting‌ increase in data volumes.

2.2 Team vision and‌ approach

We firmly believe the success of the‌ IoT lies in: (i) the network design choices‌ determining how devices are integrated, (ii) the intelligence‌ of algorithms, protocols, and services accurately interpreting demands‌ and purposes, and (iii) the adaptability of the‌ device-edge-core communication loop enabling rapid responses and efficient‌ network management. Hence, we base our approach on‌ the combination of data or communication learning, modeling,‌ algorithms/protocols design, and experimental research, while meeting the‌ requirements and challenges brought by the IoT to‌ the edge of the Internet. Therefore, our is‌ organized around the following research directions:

Technologies for‌ accommodating low-end IoT devices (resource-limited IoTs): We tackle‌ the optimization, simplification, and unification requirements imposed by‌ the heterogeneity and low capabilities of low-end IoT‌ devices. This brings the necessity to deal with‌ limitations and to propose solutions close to hardware‌ and software specifications.
Technologies for leveraging high-end IoT‌ devices' advent (smart IoTs): We focus on learning‌ the behaviors of high-end IoT devices, the smart‌ devices. The idea is to take advantage of‌ the "how in the spatiotemporal scale"‌ and the "for what purpose" these‌ devices use the network resources. This brings the‌ human element into play, in which dynamics and‌ uncertainties are shaping the way their devices interact‌ with the edge of the Internet and, consequently,‌ request and consume network resources and services.
Technologies‌ for edge-core network interaction: This element closes the‌ network $\leftrightarrow$ usability $\leftrightarrow$ device $\leftrightarrow$ network loop"‌ by bringing solutions supporting functions and communication between‌ IoT devices and the core of the Internet‌ while putting into practice the solutions proposed in‌ the two previous directions.

Through these three research‌ axes, the team places its efforts on the‌ three main elements composing the ecosystem of IoT‌ devices: (1) the device itself, (2) their usability,‌ and (3) their network context. Together, these research‌ axes will contribute to our vision toward a‌ Smart, Unified, and Tactful Internet edge skilled for‌ answering the application, services, or end-users' purposes.

3‌ Research program

Following up on the effort initiated‌ by the team members during the last few‌ years and building on an approach combining protocol‌ design, data analytics, and experimental research, we propose‌ a research program organized around three closely related‌ objectives that are briefly described in the following.‌

[Axis 1] Technologies for accommodating low-end IoT devices:‌ The IoT connects billions of low-end devices to‌ the Internet, and billions more are projected, significantly‌ increasing machine-to-machine communication. Unlike high-end devices based on‌ microprocessors, low-end IoT devices are based on‌ microcontrollers, highly resource-constrained in energy, memory, and‌ computational power. These characteristics prevent them from running‌ standard platforms like Linux or complex protocol stacks‌ based on TCP/IP. Addressing these challenges requires: (i)‌ optimized communication protocols that align with evolving radio technologies and device constraints;‌ (ii) tailored software platforms‌ offering modular updates, high-level‌‌ programming, machine learning support and energy efficiency features;‌ (iii) unification of fragmented‌ low-end IoT technologies to‌‌ ensure seamless integration with core and edge networks,‌ and (iv) low-footprint cybersecurity‌ mechanisms which enable such‌‌ devices to take part in cyberphysical, ditributed systems‌ without becoming the weakest‌ link. To support these‌‌ advancements, we propose targeted research activities addressing wireless‌ communication evolution for constrained‌ IoT devices.
[Axis 2]‌‌ Technologies for leveraging high-end IoT devices' advents: Our‌ reliance on pervasive connectivity‌ and extensive usability of‌‌ high-end IoT devices allows capturing human life patterns‌ of end-users printed in‌ the digital world. Thus,‌‌ human dynamics directly influence how resources, services, and‌ infrastructures are utilized at‌ the Internet Edge, shaping‌‌ where, when, how, and what is accessed. Consequently,‌ studying end-users' behavioral patterns‌ (e.g., mobility, content preferences)‌‌ and incorporating the inherent heterogeneity and unpredictability into‌ networking solutions is critical.‌ This challenge underpins Axis‌‌ 2, which aims to establish a tactful networking‌ design practice – enabling‌ networks to observe, interpret,‌‌ and adapt dynamically to the daily life features‌ of high-end IoT devices'‌ end-users. The research activities‌‌ of this axis focus on three main topics:‌ (i) extracting high-end IoT‌ footprints in networking data‌‌, while enforcing data representativeness and trajectories inference;‌ (ii) end-users' patterns understanding‌ at the Internet edge‌‌, for profiling and prediction of individual spatio-temporal‌ usability of the Internet‌ edge, emphasizing novelty- and‌‌ routine-like mobility modeling and urban flow understanding; and‌ (iii) addressing vulnerability and‌ security concerns linked to‌‌ end-users' patterns in digital datasets.
[Axis 3] Articulating‌ the IoT edge with‌ the core of the‌‌ network: The edge acts as the interface between‌ IoT devices and the‌ core network, addressing interoperability,‌‌ heterogeneity, and mobility. It supports several intermediary functions‌ when connecting devices to‌ the Internet. Our work‌‌ in this axis, more so than in the‌ other axis, proceeds on‌ three distinct levels: the‌‌ first level is the application area (e.g., UAV,‌ V2X, generic Edge/Cloud), the‌ second level is the‌‌ underlying technologies (e.g., blockchains, information-centric networking), and the‌ third level is the‌ specific methods and techniques‌‌ (e.g., AI/ML, RL, federated learning, split computing, offloading).‌ Each study generally combines‌ two or three of‌‌ these levels. We present them according to the‌ level we emphasize:, heterogeneity,‌ and mobility challenges (i)‌‌ Decentralized network mechanisms and architectures (application area: V2X),‌ (ii) Machine Learning enhanced‌ network protocols and classical‌‌ network optimization (methods and techniques), (iii) Edge network‌ offloading (methods and techniques),‌ (iv) Security of the‌‌ edge/core compound including IoT deployments technologies).

4 Application‌ domains

Hereafter, we describe‌ the general 1) domains‌‌ of research of TRiBE and 2) the contexts‌ and applications where our‌ solutions are applied.

Research‌‌ domains: Our research spans a range of domains,‌ including computer science, mobile‌ wireless networks, Internet of‌‌ Things (IoT), Tactile Internet, human mobility analytics and‌ prediction, edge-smart resource allocation,‌ IoT software design, social‌‌ networks, energy-saving technologies, network‌ security and user privacy, and mobility-aware networking solutions.‌
Tactile Internet: As the next evolution of the‌ IoT, the Tactile Internet focuses on applications combining‌ ultra-low latency with extremely high availability, reliability, and‌ security. These applications demand smart interactions between individuals‌ and devices, as well as device-to-device communication, enabling‌ real-time and reliable interactions for industrial, societal, and‌ business use cases. Examples of applications include automation‌ and smart transportation, 3D and educational games, and‌ x-reality applications and services. Our solutions aim to:‌ (i) introduce intelligence and quasi-in-time adaptivity (accounting for‌ individuals' behaviors, IoT limitations, and the context of‌ services and environments) in networking's resource allocation, management,‌ and usability, and (ii) contribute to achieving the‌ goals of the Tactile Internet.
IoT twins: The‌ rise of IoT is driving the emergence of‌ digital twins—digital copies of complex systems—operable via 5G‌ or next-generation networks. Digital twins provide real-time information‌ on working/leisure areas, traffic, weather conditions, air quality,‌ and more, for a city, region, or even‌ an entire country. These highly distributed systems require‌ robust and reactive blockchain mechanisms to manage massive‌ data flows from millions of sensors transmitting at‌ high frequency. Our solutions in IoT-massive edge applications,‌ analytic learning theory, and frugal AI are designed‌ to support the development of such systems. Additionally,‌ digital twins for traffic, transportation, and geographic visitation‌ will require insights into population mobility and space‌ usage. Our solutions in mobility understanding, profiling, and‌ simulation are well-suited to advance these applications.
Urban‌ planning and disaster management applications: Our research provides‌ critical insights for applications in urban planning and‌ disaster management. For instance, topics such as SafeCityMap:‌ From spatiotemporal mobility of our society to COVID‌ propagation understanding and Geometry of virus exposure (detailed‌ in the "New Results" section) exemplify our work.‌ Broadly speaking, mobility analytics from these projects enhance‌ understanding of urban space usage and support: (i)‌ Epidemic prevention and disaster response; (ii) Urban traffic‌ management and mobility prediction algorithms; (iii) Provision of‌ energy-efficient, cost-effective network infrastructures tailored to changing mobility‌ patterns. Our tools and investigations also enable the‌ study of spatiotemporal activity in geographic areas (e.g.,‌ visitation rates and patterns), providing insights into the‌ socio-economic impacts on residential or activity-based zones—particularly during‌ situations like lockdown periods.
Network security, privacy, and‌ adversarial threats: Modern communication networks are increasingly exposed‌ to large-scale security and privacy threats driven by‌ their growing complexity, automation, and reliance on data-driven‌ mechanisms. Such threats include telecom fraud (e.g., SIMBox‌ and bypass fraud), malicious activities at the cellular‌ and network edge, and large-scale social engineering attacks‌ targeting end users. Addressing these challenges requires fine-grained‌ visibility into network behaviors and user interactions, as‌ well as models capable of capturing adaptive and‌ adversarial strategies. Our contributions rely on data-driven and‌ machine-learning-based methods for behavioral analysis, attack detection, and‌ anomaly identification, leveraging detailed mobile network data (e.g.,‌ signaling, CDR/XDR) and edge-level observations. Particular attention is‌ given to adversarial machine learning and privacy-aware analysis, with the objective of‌ designing robust defenses that‌ strengthen network resilience while‌‌ preserving user data privacy.
Additional applications influenced by‌ our research are discussed‌ in the next section‌‌ on Social and Environmental Responsibility.

5 Social and‌ environmental responsibility

5.1 Footprint‌ of research activities

We‌‌ believe our research can benefit society and the‌ environment because:

The Internet‌ of Things (IoT) is‌‌ set to dramatically increase the number of connected‌ devices, potentially raising network‌ power consumption and environmental‌‌ impact. However, many IoT applications address environmental management‌ by monitoring and resolving‌ critical issues. Most devices‌‌ are low-power wireless systems, often solar-powered. Our research‌ focuses on optimizing efficient‌ low-end networks and minimizing‌‌ the costs of creating sensor field digital twins‌ through green blockchain designs.‌
Smart devices, inheriting user‌‌ dynamics and decision-making, introduce uncertainties in predicting where‌ and when network resources‌ are needed. The common‌‌ Internet response has been over-provisioning resources to manage‌ this instability, which also‌ exacerbates energy inefficiency. However,‌‌ in a predominantly mobile Internet, such practices inflate‌ energy consumption, becoming both‌ costly and unsustainable, necessitating‌‌ a strategic re-evaluation. Our research fosters just-in-time networking‌ resource usability.
The European‌ Commission’s Sustainable and Smart‌‌ Mobility Strategy (2020) underscores that achieving the Green‌ Deal’s goals hinges on‌ creating a sustainable transport‌‌ system. This transformative vision reshapes how we view‌ transport usability and availability,‌ human mobility, and its‌‌ interaction with spatial dynamics, emphasizing the need to‌ understand mobility behavior and‌ its drivers of change.‌‌ Our research emphasizes such aspects.

The previous assertions‌ naturally guide our research‌ and envisioned outcomes. TRiBE's‌‌ research contributes to environmental and societal responsibilities in‌ the following ways:

TRiBE‌ research is targeting a‌‌ network intelligence much closer to end-users – and‌ consequently, to the edge‌ of the Internet. In‌‌ this sense, edge intelligence (i.e., learning, reasoning, and‌ decision-making) provides distributed autonomy,‌ replacing the classical centralized‌‌ structures. TRiBE results thus, contribute to (1) smartly‌ using networking resources, (2)‌ using a lower amount‌‌ of aggregated power in dispersed locations, and (3)‌ avoiding the energy consumption‌ related to the transmission‌‌ of information back and forth to the Internet‌ core. This conviction is‌ the common thread in‌‌ the suitable by-design solutions of the 2nd and‌ 3rd TRiBE's axis/focus,‌ which will naturally contribute‌‌ to the new energy-efficient architectural evolution of the‌ Internet.
TRiBE research pursues‌ the conviction that methods‌‌ allowing to smartly and efficiently allocate/use resources (of‌ devices and the network)‌ at the Internet edge‌‌ are energy-friendly and contribute to the IT sector's‌ electricity needs. This conviction‌ is also the common‌‌ thread behind the 2nd and 3rd TRiBE's axis‌.
Besides, the understanding‌ of the way carried‌‌ high-end IoT devices move and interact with one‌ another (i.e., related to‌ axis 2 and 3‌‌ of TRiBE) has the potential to impact epidemiology‌ studies, urbanization investigation, and‌ Internet provisioning (e.g., in‌‌ the successful comprehension of the spread of epidemics‌ or of the population;‌ in urban planning; in‌‌ intelligent transportation systems in‌ smart cities; for urban space management; or in‌ more suitable-for-devices resource allocation. The SafeCityMap 62 and‌ Ariadne Covid Inria-Covid projects carried by members of‌ the team reinforce such assertion. In particular, the‌ SafeCityMap project investigates the impact of the 1st,‌ 2nd, and 3rd lockdown on the regular mobility‌ habits of the Paris population. Results of such‌ investigations are posted in the interactive SafeCityMap website‌. Besides, our recent investigation shows a natural‌ correlation between pollution indicators and SafeCityMap results describing‌ mobility preferences and landscape usability in Paris: Indicators‌ having the potential to impact society and population‌ health.
In the 1st TRiBE's axis,‌ TRiBE goals also relate to the provision of‌ optimized communication protocols and software solutions designed to‌ fit the stark specificities of low-end IoT devices‌ while taking into account radio technology evolution. The‌ motivation here is to efficiently use and manage‌ the billions of low-end devices expected to (i)‌ gradually connect to and (2) drastically increase the‌ communication, and consequently, the energy consumption, on the‌ Internet. TRiBE's 1st research axis pursues the conviction‌ that the smart accommodation of low-end IoT devices'‌ related solutions will contribute to energy efficiency at‌ the Internet edge. In a part of our‌ research work, we focus on constrained devices (constrained‌ in processing power and energy) and provide efficient‌ algorithms in computation and communication reduction, both being‌ translated into energy savings, reducing, thus, the energy‌ footprint of the IoT.
A sizable part of‌ our research activities is carried on top of‌ open-source software that we develop, and especially the‌ open source software platforms RIOT and Ariel OS‌, which are operating systems for the Internet‌ of Things, targeting low-power embedded devices based on‌ microcontrollers (i.e., related to axis 1 of TRiBE‌). In this way, research and developments that‌ improve energy efficiency directly on RIOT or Ariel‌ OS are made readily available to IoT practitioners.‌ Several TRiBE members contribute actively to this platform,‌ around which a large international community has snowballed.‌ In this way, research and developments that improve‌ energy efficiency are made readily available to IoT‌ practitioners, e.g., through RIOT, Ariel OS or other‌ software in the ecosystem.

5.2 Impact of research‌ results

Ethics: We handle ethical issues in our‌ research as developed in Scientific integrity and open‌ science practices (see subsections 4.1 and 4.2).

Socio-economic‌ impact:

As a result of TRiBE engagement in‌ EU environment and green priorities, the team is‌ strongly involved in four projects of two national‌ research actions (PEPR): (1) the MOBIDEC‌ (i.e., Digitalisation and Decarbonisation of Mobilities) focusing‌ on the digital and carbon neutrality of mobility‌ and (2) the 5G and Network of Future‌ aiming the development of 5G and 6G while‌ assessing their environmental impacts. Both PEPRs and, consequently,‌ the contributions of the team on such actions‌ will contribute to make research impacting the environment‌ and society while ensuring the security of transmitted data and privacy compliance‌ of treated mobility traces.‌ This engagement is present‌‌ in all the three TRiBE's research axes.
When‌ privacy concerns are identified,‌ TRiBE has dedicated efforts‌‌ in:
- designing solutions to ensure anonymization and/or fraud‌ detection of wireless networks'‌ datasets. Related to‌‌ the anonymization concern, we point out important privacy-related‌ flaws in current wireless‌ communication standards 64.‌‌ Our related designed solutions highlight the possibility to‌ efficiently (i) identify devices‌ associated with randomized addresses‌‌ and (ii) reconstruct their trajectories only based on‌ signal measurements (cf. the‌ PhD thesis of Abhishek‌‌ Kumar Mishra 66).
- quantifying privacy risks in‌ mobility datasets through behavioral‌ exposure analysis. By‌‌ going beyond location- and trajectory-based notions of exposure,‌ this work enables interpretable,‌ data-driven assessment of user‌‌ privacy risks and supports the identification of vulnerable‌ individuals and usage patterns‌ that remain invisible to‌‌ existing approaches 27, 40, 26.‌ These insights pave the‌ way for tailored and‌‌ cost-efficient privacy protection strategies when sharing or releasing‌ mobility datasets, while better‌ preserving data utility.
Besides,‌‌ the team contributions on cellular fraud detection address‌ the societal and economic‌ impact of SIMBox-based international‌‌ bypass fraud in cellular networks. Our work supports‌ the development of practical,‌ edge-based detection mechanisms suitable‌‌ for operational deployment by network operators 31.‌ In 2025, these results‌ led to concrete valorisation‌‌ actions, including engagement with the Inria Startup Studio‌ to explore industrial transfer‌ opportunities.
Our work on‌‌ embedded software platforms (RIOT and Ariel OS) impacts‌ the industry. Products ship‌ with RIOT since 2017,‌‌ and several companies are currently developing prototypes using‌ Ariel OS. The open‌ source nature of these‌‌ platforms, combined with the strong cybersecurity functionalities they‌ provide (e.g. secure software‌ update over the low-power‌‌ networks) makes these platforms appealing as basis for‌ small, medium and large‌ companies in various verticals.‌‌
Other contributions such as 67, 65,‌ 60, 58 demostrate‌ the engagement of the‌‌ team in enforcing the carbon neutrality and the‌ green management of mobility‌.
Last but not‌‌ least, another means for our research results to‌ have an impact is‌ through contributions to standardization‌‌ (including IETF): TRiBE members co-author standards and help‌ to define and specify‌ efficient protocols and their‌‌ optimization.

General audience: We have also been intervening‌ in the public debate‌ and fostering science disseminaiton:‌‌

Anne Josiane Kouam co-organized a one-week mathematics and‌ computer science school in‌ Yaoundé (Cameroon) within the‌‌ international programs of Animath and PromoMaths (August 2025).‌ She delivered intensive teaching‌ sessions (20 hours) covering‌‌ mobile data privacy, statistics, and introductory machine learning,‌ and participated in the‌ Miss STEM Cameroon initiative‌‌ to promote scientific engagement and inclusion among high-school‌ girls.
Geoffrey Deperle took‌ part in several scientific‌‌ mediation activities targeting high-school students. As part of‌ the program “1 scientifique,‌ 1 classe : Chiche!”‌‌, he presented research careers at Lycée Les‌ 7 Mares (Maurepas), aiming‌ to challenge stereotypes and‌‌ highlight the diversity of‌ academic paths. He also contributed to an observation‌ internship for seconde students by introducing the TRiBE‌ team and research activities, followed by a hands-on‌ session on Python-based fractal design to illustrate geometric‌ transformations and computational thinking.

6 Highlights of the‌ year

This section reports major scientific distinctions and‌ achievements of the TRiBE project-team in 2025 that‌ had a significant impact at the national or‌ international level.

6.1 Awards

National Award for Open‌ Science. Anne Josiane Kouam, starting researcher in the‌ TRiBE project-team, received the ”Science Ouverte de la‌ Thèse” 2025 Award from the French Ministry of‌ Higher Education and Research. This national distinction recognizes‌ the exemplary open-science contribution of her doctoral work,‌ which addresses a large-scale yet poorly understood phenomenon‌ in telecommunications networks: international bypass fraud, commonly known‌ as SIMBox fraud.
Best Paper Award at an‌ A-ranked International Conference. The paper entitled “Beyond Aggregates:‌ A Fine-Grained Analysis of Individual Mobility and Traffic‌ Dependencies” 32, authored by Anne Josiane Kouam‌ et al., received the Best Paper Award‌ at the 27th International Conference on Modeling, Analysis‌ and Simulation of Wireless and Mobile Systems (ACM‌ MSWiM 2025), an A-ranked international conference. The paper‌ was presented by Anne Josiane Kouam and provides‌ an original individual-level analysis of mobility and traffic‌ dependencies using large-scale mobile network data.

6.2 Scientific‌ Events and Conferences

Co-organization of a Flagship International‌ Conference. The TRiBE project-team co-organized the 2025 edition‌ NetMob conference, held at CNAM in Paris from‌ October 8 to October 10, 2025. NetMob is‌ the leading interdisciplinary conference dedicated to the analysis‌ of mobile phone and mobility data, at the‌ intersection of data science, network science, social sciences,‌ and industrial applications. The 2025 edition gathered 155‌ participants from 29 countries across six continents, including‌ representatives from academia, public institutions (e.g., Inserm, UNICEF),‌ and industry. The conference was organized in collaboration‌ with CNAM, IFPEN, and TU Berlin, and supported‌ by major sponsors including Inria, PEPR MOBIDEC, Qualcomm,‌ and Orange. On this occasion, the TRiBE team‌ also launched a new edition of the NetMob‌ Data Challenge, leveraging real mobility datasets provided by‌ the PEPR MOBIDEC MobSci Dat-Factory.
Co-organization of PEMWN‌ 2025. Members of the TRiBE project-team also co-organized‌ IEEE/IFIP PEMWN 2025, the 14th IFIP International‌ Conference on Performance Evaluation and Modeling in Wired‌ and Wireless Networks. It is a scientific forum‌ for researchers to present and learn significant contributions‌ and interesting ideas on a wide range of‌ research topics on networking and closely related areas.‌ It was held at CNAM, November 25-27, 2025,‌ and was co-organized with University of Rouen, ENSI/University‌ of Manouba, CNAM, UCLM (Spain), ETS Montreal, ENIT,‌ Sorbonne University.

6.3 Books

Publication of a Research‌ Monograph in an International Reference Series. Philippe Jacquet‌ authored the book ”Paradoxes and Physical Limits of‌ Information Theory” 63, published in April 2025‌ as Volume 1 of the World Scientific Series‌ on Quantum Algorithms, Information, and Learning. The book provides a rigorous‌ theoretical investigation of paradoxes‌ and physical constraints in‌‌ information theory, with implications for quantum information processing‌ and learning systems.

7‌ Latest software developments, platforms,‌‌ open data

Our team emphasizes real-world implementation of‌ research contributions, ensuring theoretical‌ findings are validated through‌‌ practical development. Each project includes a functional prototype,‌ framework, or software tool,‌ developed iteratively alongside research‌‌ advancements. Committed to privacy and ethics, the team‌ ensures developed software adhere‌ to these principles. Following‌‌ an open-source approach, all software and frameworks are‌ made publicly available under‌ open-source licenses whenever possible.‌‌ Moreover, even though part of our work involves‌ collecting or handling private‌ data, we are committed‌‌ to ensuring that our tools respect privacy and‌ adhere to ethical standards.‌

7.1 Latest software developments‌‌

7.1.1 RIOT

Name:
RIOT
Keywords:
Internet of things,‌ Operating system, Sensors, Iot,‌ Wireless Sensor Networks, Internet‌‌ protocols
Scientific Description:
While requiring as low as‌ 1,5kB of RAM and‌ 5kB or ROM, RIOT‌‌ offers real time and energy efficiency capabilities, as‌ well as a single‌ API (partially POSIX compliant)‌‌ across heterogeneous 8-bit, 16-bit and 32-bit low-hardware. This‌ API is developer-friendly in‌ that it enables multi-threading,‌‌ standard C and C++ application programming and the‌ use of standard debugging‌ tools (which was not‌‌ possible so far for embedded programming). On top‌ of this, RIOT includes‌ several network stacks, such‌‌ as a standard IPv6/6LoWPAN stack and a information-centric‌ network stack (based on‌ CCN).
Functional Description:

RIOT‌‌ is an Open Source operating system that provides‌ standard protocols for embedded‌ systems. RIOT allows, for‌‌ example, the development of applications that collect sensor‌ data and transmit it‌ to a central node‌‌ (e.g. a server). This data can then be‌ used for smart energy‌ management for instance.

RIOT‌‌ is specially designed for embedded systems, which are‌ strongly constrained in memory‌ and energy. Further, RIOT‌‌ can easily be ported to different hardware devices‌ and follows the latest‌ evolution of IP standards.‌‌

RIOT applications can readily be tested in the‌ FIT IoT-Lab, which provides‌ a large-scale infrastructure facility‌‌ with 3000 nodes for testing remotely small wireless‌ devices.
News of the‌ Year:

4 releases in‌‌ 2025: 2025.10 2025.07 2025.04 2025.01

Last 2025.10 release‌ includes:

- The new‌ XIPFS filesystem module allows‌‌ executing code from memory-mapped flash on supported platforms.‌

- Support for the‌ ESP32-C6 RISC-V WiFi MCU‌‌ was added.

- Nearly all of the guides‌ were migrated, adapted or‌ rewritten to the new‌‌ guide site.

- You can now also read‌ all RIOT release notes‌ directly on the guide‌‌ site, outside of Github.

- NanoCoAP server will‌ no longer copy requests‌ to an internal buffer‌‌ for processing, instead it operates directly on the‌ network buffer, eliminating the‌ case of 'too large'‌‌ requests.

- The new hosts module allows to‌ statically assign hostnames to‌ IP addresses, similar to‌‌ /etc/hosts on UNIX.

- The u8g2 package now‌ implements the disp_dev, meaning‌ e.g. LVGL can now‌‌ be used on all‌ displays supported by u8g2.

- Support for more‌ members of the STM32C0 family was added.

-‌ Initial support for the RP2350.

Statistics for 2025‌ (lower bounds): - 180 pull requests - 411‌ commits, have been merged since the last release‌ - 16 issues have been solved. - 31‌ people contributed with code in 121 days -‌ 3372 files have been touched with 51234 (+)‌ insertions and 205847 deletions (-).
URL:
http://www.riot-os.org
Contact:‌
Emmanuel Baccelli
Participants:
Emmanuel Baccelli, Koen Zandberg, Oliver‌ Hahm, Francois-Xavier Molina, Alexandre Abadie
Partners:
Freie Universität‌ Berlin, University of Hamburg

7.1.2 FraudZen

Keywords:
Fraud‌ detection, SIMBox fraud, LTE, Simulator
Scientific Description:
FraudZen‌ is an open-source simulator of the activities (traffic‌ and mobility) and interactions of legitimate and SIMBox‌ fraudulent users, on the top of a realistic‌ cellular network infrastructure. From input models of legitimate‌ and fraudulent behaviors, FraudZen generates CDRs datasets.
Functional‌ Description:

FraudZen is an open-source simulator of SIMBox‌ fraud strategies in LTE networks. It is designed‌ to tackle the lack of fraudulent and up-to-date‌ CDRs, which is the ground truth required for‌ efficient SIMBox fraud mitigation.

FraudZen reproduces the realistic‌ cellular network architecture of a SIMBox fraud’s target‌ area and simulates the network usage and interactions‌ of legitimate and SIMBox fraudulent users on top‌ of this architecture. FraudZen’s resulting CDRs convey users’‌ communication behavior at individual fine-grained precision. Researchers and‌ mobile operators can use this tool to (i)‌ inject fraudulent traffic to their CDRs and check‌ the validity of their designed solutions, (ii) analyze‌ the impact of the so-farunreachable SIMBox ecosystem, i.e.,‌ SIMBox architecture and fraud parameters, (iii) reproduce and‌ explore off-net fraud mechanisms, and (iv) design and‌ investigate new fraud schemes. The full control and‌ flexibility related to the simulation environment guarantee complete‌ and large fraudulent CDRs ground truth for detection‌ models’ training. Moreover, FraudZen allows anticipating the fraud‌ evolution, freeing research from the past/current fraud capabilities‌ and allowing the incorporation of not-yet-existing SIMBox functionalities‌ in foresight.
News of the Year:

In 2025:‌

- Simulator extention through the public release of‌ the FraudZen Dataset: Realistic Ground Truth CDRs of‌ Bypass Fraud Techniques in Mobile Networks on the‌ national open data platform recherche.data.gouv.fr (https://entrepot.recherche.data.gouv.fr/dataset.xhtml?persistentId=doi:10.57745/TAV6PQ).

- Dataset‌ provides synthetic call-detail records generated by the FraudZen‌ simulator, modeling both legitimate mobile subscribers and SIMBox‌ fraudsters under a wide range of threat strategies,‌ including mobility-, traffic-, and social-based fraud scenarios.

->‌ It offers a realistic and privacy-preserving benchmark for‌ evaluating fraud- and anomaly-detection techniques in cellular networks,‌ supporting reproducible research and technology transfer in telecom‌ security.

-> It strengthens FraudZen’s role as an‌ open experimentation framework, complementing ongoing research, hackathon activities,‌ and future integration of mature detection mechanisms into‌ the simulator.
URL:
https://gitlab.inria.fr/simbox-fraud-mitigation/fraudzen
Publications:
hal-03838853, hal-04543435‌
Contact:
Anne Josiane Kouam Djuigne
Participants:
Anne Josiane‌ Kouam Djuigne, Aline Carneiro Viana, Alain Tchana

7.1.3‌ MITIK-MGMT

Name:
MITIK Data Collector Management Tools
Keywords:‌
Wi-Fi, Infrastructure software, Mobile Crowdsensing
Scientific Description:
MITIK-MGMT is an open-source management‌ tool developed as part‌ of the MITIK project‌‌ and aims to automate the configuration process and‌ management of experiments using‌ WiFi collectors offered in‌‌ MITIK. The supported functions are: - Provide a‌ tool that allows the‌ simultaneous configuration of multiple‌‌ collectors. - Centralized management of several collectors (synchronization,‌ raw data capture, data‌ transfer, and processing...). -‌‌ Configuration of parameters and execution of MITIK project‌ modules.
Functional Description:
The‌ objective of the MITIK‌‌ project is to carry out non-intrusive passive measurements‌ to analyze the mobility‌ of users through contacts‌‌ during their travels. The objective is to use‌ probe-request packets coming from‌ mobile devices using WiFi-type‌‌ wireless communications. MITIK-MGMT is a management tool developed‌ as part of the‌ MITIK project and enables,‌‌ through a "sniffer manager," the automated setup and‌ management of practical experiments‌ using WiFi collectors.
News‌‌ of the Year:
- Maturated version of the‌ MITIK-MGMT tool, with revisions‌ of documentation and parametrization.‌‌ - The code was also recently deposited as‌ open-source with GPLv3.0-or-later license.‌ - The code is‌‌ registered at Software Heritage as referred in document‌ hal-04814847 in HAL. -‌ The following HAL report‌‌ provides configuration and usability instructions: https://inria.hal.science/hal-04818320v1 - For‌ dependencies among MITIK tools,‌ refer to MITIK-GUIDE (https://gitlab.inria.fr/mitik/mitik-guide)‌‌
URL:
https://gitlab.inria.fr/mitik/measurement-management/mitik-mgmt
Publications:
hal-04568193, hal-04814847
Contact:
Nadjib‌ Achir
Participants:
Aline Carneiro‌ Viana, Nadjib Achir, Fernando‌‌ Molano Ortiz, Fernando Dias De Mello Silva

7.1.4‌ MITIK-SENS

Name:
Privacy-preserving WiFi‌ Sniffer tool
Keywords:
Wi-Fi,‌‌ Privacy
Scientific Description:
Public wifi (IEEE 802.11) networks‌ are an abundant data‌ source that may serve‌‌ different applications such as epidemic tracking and prevention,‌ disaster response, crowdsensing, or‌ ubiquitous urban services. Nevertheless,‌‌ collecting and exploiting such data brings many privacy‌ liabilities, considering that each‌ transmitted frame has the‌‌ MAC address (a unique device identifier) of the‌ corresponding personal device, also‌ considered sensitive information. Literature‌‌ has shown that the MAC randomization performed by‌ phone manufacturers is insufficient‌ to protect devices' identification.‌‌ Data obfuscation is a promising solution to avoid‌ storing advertised identifiers of‌ devices and prevent attackers‌‌ from acquiring sensitive data. Obfuscating such identifiers while‌ also being able to‌ differentiate frames sent by‌‌ different devices poses a significant challenge for frame‌ capturing by low-resource IoT‌ devices in real time.‌‌ Since no popular off-the-shelf sniffer (wireshark or tcpdump,‌ etc..) allows for on-the-fly‌ obfuscation, we build a‌‌ new custom-made sniffer module **MITIK-SENS** capable of on-the-fly‌ obfuscating (hash and truncate)‌ the required data needed‌‌ of each wifi frame to protect user privacy.‌
Functional Description:
Privacy-preserving WiFi‌ Sniffer tool with on-the-fly‌‌ MAC Address Obfuscation.
News of the Year:
-‌ 1st version of the‌ MITIK-SENS tool, with revisions‌‌ of documentation and parametrization. - The code was‌ also recently deposited as‌ open-source with GPLv3.0-or-later license‌‌ and registered at Software Heritage, as referred in‌ doucment hal-04816385 in HAL.‌ - The following HAL‌‌ report provides configuration and usability instructions: https://inria.hal.science/hal-04818079v1 -‌ For dependencies among MITIK‌ tools, refer to MITIK-GUIDE‌‌ (https://gitlab.inria.fr/mitik/mitik-guide)
URL:
https://gitlab.inria.fr/mitik/anonymous-measurement/mitik-sens
Publications:‌
hal-03906600, hal-04568193, hal-04816385
Contact:
Fernando Dias‌ De Mello Silva
Participants:
Aline Carneiro Viana, Nadjib‌ Achir, Luis Henrique Maciel Kosmalski Costa, Fernando Molano‌ Ortiz, Fernando Dias De Mello Silva, Anne Fladenmuller,‌ Abhishek Mishra

7.1.5 MITIK-LINK

Name:
MITIK MAC Address‌ Association
Keywords:
Wi-Fi, Probe-requests, MAC address randomization, Frame‌ association
Scientific Description:
MITIK-LINK is a tool designed‌ to associate randomized MAC addresses within WiFi network‌ traces gathered from the MITIK-SENS tool.
Functional Description:‌
MITIK-LINK performs the MAC association of randomized MAC‌ addresses used by the same device. This tool‌ models the frame association to resolve MAC conflicts‌ in small intervals. It uses time and frame‌ content-based signatures to resolve and associate MACs inside‌ a conflict. Finally, a logistic regression-based algorithm using‌ the obtained signatures is proposed to associate devices‌ with similar signatures.
News of the Year:
-‌ 1st version of the MITIK-LINK tool, with revisions‌ of documentation and parametrization. - The code was‌ also recently deposited as open-source with GPLv3.0-or-later license‌ and registered at Software Heritage, as referred in‌ Document hal-04815312 in HAL. - The following HAL‌ report provides configuration and usability instructions: https://inria.hal.science/hal-04818359v1 -‌ For dependencies among MITIK tools, refer to MITIK-GUIDE‌ (https://gitlab.inria.fr/mitik/mitik-guide)
URL:
https://gitlab.inria.fr/mitik/mac-association/mitik-link
Publications:
hal-04815312, hal-03298339,‌ hal-04171864, tel-04311364, hal-04568193
Contact:
Nadjib Achir‌
Participants:
Abhishek Mishra, Aline Carneiro Viana, Nadjib Achir,‌ Fernando Molano Ortiz

7.1.6 MITIK-TRAJ

Name:
MITIK-TRAJ -‌ WiFi devices trajectory inference tool
Keywords:
Wi-Fi, Trajectory‌ Generation, Mobility
Scientific Description:
MITIK-TRAJ is a tool‌ for trajectory reconstruction of a WiFi mobile terminal.‌ It leverages the signal strength of users' public‌ WiFi probe requests collected from measurements of multiple‌ deployed or sniffers. Characterize and approximate the error‌ in the radial distances between the device and‌ the sniffer. Leverage the error characterization and approximated‌ radial distances to estimate the bounds associated with‌ a device's position. Finally, considering the spatiotemporal bounds‌ of device positions over time, it infers the‌ user's bounded trajectory.
Functional Description:
MITIK-TRAJ is a‌ tool for reconstructing the trajectory of equipment from‌ their Wi-Fi traces by introducing the concept of‌ bounded trajectory. The tool considers three significant components:‌ i) Generating observation sets, ii) Characterizing radial-distance estimation‌ errors, and iii) Obtaining bounded trajectories.
News of‌ the Year:
- 1st version of the MITIK-TRAJ‌ tool, with revisions of documentation and parametrization. -‌ The code was also recently deposited as open-source‌ with GPLv3.0-or-later license and registered at Software Heritage,‌ as referred in Document hal-04924988 in HAL. -‌ The following HAL report provides configuration and usability‌ instructions: https://inria.hal.science/hal-04925002v1 - For dependencies among MITIK tools,‌ refer to MITIK-GUIDE (https://gitlab.inria.fr/mitik/mitik-guide)
URL:
https://gitlab.inria.fr/mitik/trajectory-reconstruction/mitik-traj
Publications:
hal-03906579‌, tel-04311364, hal-04568193, hal-04924988
Contact:
Nadjib‌ Achir
Participants:
Abhishek Mishra, Aline Carneiro Viana, Nadjib‌ Achir, Fernando Molano Ortiz

7.1.7 MobilityPulse

Name:
MobilityPulse‌ - Routine and Novelty-Seeking Behaviors Analysis Framework
Keywords:‌
Mobile phone, Human mobility, Mobility, Behavior modeling, Statistical‌ analysis, Profiling, Prediction, Predictive analytics
Scientific Description:

This‌ framework examines routine and exploratory tendencies in human mobility, influencing mobility predictability‌ and practical visit predictions.‌ On the routine side,‌‌ it extracts the predictability of input datasets and‌ identifies three key mobility‌ features—regularity, stationarity, and diversity—that‌‌ impact predictability. Additionally, it facilitates the analysis of‌ contextual factors affecting predictability.‌

From the exploratory perspective,‌‌ existing mobility research struggles to accurately capture novelties‌ in human movement, where‌ the severity of uncertainty‌‌ influences prediction accuracy. This framework introduces a two-dimensional‌ mobility model that explicitly‌ accounts for regular and‌‌ exploratory behaviors. It also enables individuals to be‌ classified into three mobility‌ profiles: Scouters, Routiners, and‌‌ Regulars. Moreover, we analyze the mobility features of‌ each profile— e.g., regularity,‌ radius of gyration (RoG),‌‌ stationarity, diversity, maximum displacement, etc — to characterize‌ mobility behavior. Furthermore, the‌ framework comprehensively evaluates how‌‌ novelty-seeking tendencies affect theoretical and practical mobility predictability.‌
Functional Description:
MobilityPulse suggests‌ focusing on the "heartbeat"‌‌ or rhythm of human movement, capturing both the‌ regular patterns (routine) and‌ the spikes of activity‌‌ that represent novelty-seeking behaviors. The framework analyzes the‌ underlying dynamics (related to‌ routine-like mobility) and variations‌‌ (moments of exploration) in human mobility, much like‌ how a pulse reflects‌ changes in a person’s‌‌ physiological state. Finally, the framework allows the investigation‌ of the predictability of‌ routine-like patterns and the‌‌ ability to perform predictions while being aware of‌ moments of exploration.
Release‌ Contributions:
A significant functional‌‌ difference exists compared to the previous version, marked‌ by increased flexibility and‌ improved feature integration, beyond‌‌ documentation-related changes.
News of the Year:
- in‌ 2025, a significant functional‌ improvement characterized by greater‌‌ code flexibility, better modularity, and improved feature integration,‌ beyond the corresponding documentation-related‌ changes.
URL:
https://gitlab.inria.fr/mobility-framework/mobcomp-framework/MobilityPulse-framework
Publications:‌‌
hal-03905517, hal-02286128, hal-01367825, hal-04699871,‌ hal-02286080, hal-03360537,‌ hal-03128639, hal-02944150,‌‌ hal-03444658
Contact:
Aline Carneiro Viana
Participants:
Aline Carneiro‌ Viana, Licia Amichi, Douglas‌ Do Couto Teixeira, Jussara‌‌ Marques Almeida, Joao Paulo Esper Spindula, Antonio Alfredo‌ Ferreira Loureiro, Mark Crovella,‌ Fernando Molano Ortiz, Joao‌‌ Paulo Esper Spindula
Partners:
Inria, Federal University of‌ Minas Gerais, Boston University‌

7.1.8 En-WDM

Keywords:
Human‌‌ mobility, Simulation, Statistical modeling
Scientific Description:
En-WDM builds‌ upon WDM as its‌ foundational element, enriching its‌‌ capabilities. Our motivation to use WDM is twofold.‌ First, in contrast to‌ models found in related‌‌ literature, WDM originality comes from the combination of‌ various mobility aspects present‌ in people's daily lives‌‌ (e.g., home and workplaces, day periods). Secondly, WDM‌ closely mirrors the distributions‌ of wireless interactions, including‌‌ inter-contact and contact time, as observed in two‌ real-world measurement experiments (i.e.,‌ iMote and Dartmouth), establishing‌‌ its modeling generality. However, WDM does have limitations‌ in capturing certain nuanced‌ real mobility habits and‌‌ fine-tuning aspects. En-WDM addresses these constraints by enhancing‌ the model with additional‌ insights from the literature‌‌ on laws governing human mobility behavior. This includes‌ considerations such as preferential‌ attachment, regular daily behavior,‌‌ transportation-dependent shortest-path preferences, and, crucially, accounting for uncertainty‌ (novelty-seeking behaviors) and heterogeneity.‌ Specifically, En-WDM assigns to‌‌ the emulated users: (i)‌ Trajectories that incorporate routine- and exploration-based locations, (ii)‌ Displacement profiles, along with preferential neighborhoods (e.g., residential‌ zones, business districts), (iii) Profiles associated with the‌ maximum distance covered in their displacements and (iv)‌ Fine-grained parameterization tailored to a real-world city (i.e.,‌ Helsinki), reproduced with high fidelity.
Functional Description:

Domain-wide‌ recognized by their high value in diverse domains,‌ such as sociology, epidemiology, transportations, and networking, the‌ access to human mobility data for research faces‌ multiples challenges, related to its inherently private nature.‌

To bridge this gap, the En-WDM framework aims‌ the realistic emulation of a population urban mobility‌ in a real-world city map, with users displacements‌ generated according to public sources and describing city‌ planning and transportation information. En-WDM inherits the highly‌ configurable capability of the Opportunistic Network Environment (ONE)‌ simulator. Besides, it enhances the Working Day Mobility‌ model (WDM) of ONE into a more realistic‌ model and generates related human mobility data to‌ the format <Timestamp, userId, lat,lon>.
Release Contributions:
No‌ functional differences compared to the previous version, but‌ changes related to the documentation.
News of the‌ Year:
- Small addings in documentation and code‌ optimisation.
URL:
https://gitlab.inria.fr/simbox-fraud-mitigation/legitimate-mobile-datasets-generation/en-wdm
Publication:
hal-03910141v1
Contact:
Anne Josiane‌ Kouam Djuigne
Participants:
Anne Josiane Kouam Djuigne, Aline‌ Carneiro Viana, Alain Tchana

7.1.9 Ariel OS

Keywords:‌
Iot, Cybersecurity, Operating system, Microcontroller, Energy efficiency
Functional‌ Description:
Ariel OS is an operating system for‌ secure, memory-safe, low-power Internet of Things (IoT). It‌ is based on Rust from the ground up‌ and supports hardware based on 32-bit microcontroller architectures‌ (Cortex-M, RISC-V, and Xtensa). Ariel OS framework that‌ allows to write portable embedded Rust applications with‌ minimal boilerplate while providing a batteries-included experience.
Release‌ Contributions:
1st release of this software.
News of‌ the Year:

- First release of Ariel OS,‌ an embedded Rust library OS.

- Three releases‌ in 2025: v0.1.0, v0.2.0 and v0.2.1

- Ariel‌ OS runs on small MCUs like ARM Cortex-M,‌ Espressif Xtensa and RISC-V architectures such as nRF5x,‌ RP2xxx, STM32 and ESP32.

- Ariel OS integrated‌ Embassy HAL, a multi-core enabled preemptive scheduler, a‌ network stack and various OS services as well‌ as integrated a curated set of Rust crates‌ from the ecosystem, turning this combination into a‌ full-blown, feature-rich RTOS.
URL:
https://github.com/ariel-os/ariel-os
Publication:
hal-05119288
Contact:‌
Emmanuel Baccelli
Participants:
Emmanuel Baccelli, Romain Fouquet, Nils‌ Ponsard, Antoine Lavandier, Jeremy Kromer
Partner:
Freie Universität‌ Berlin

7.1.10 RIOT-rs

Keywords:
Microcontroller, Iot, Rust
Functional‌ Description:
RIOT-rs is an operating system for secure,‌ memory-safe, low-power Internet of Things (IoT). RIOT-rs is‌ based on Rust from the ground up. The‌ main idea is "Rust & RIOT combined for‌ ergonomic embedded development", in a nutshell: rewriting core‌ RIOT modules in Rust. Hardware targets include varieties‌ of IoT hardware based on 32-bit microcontroller architectures‌ (such as Cortex-M, RISC-V).
Contact:
Emmanuel Baccelli
Partner:‌
Freie Universität Berlin

7.1.11 MoBES

Name:
Mobility Behavior-based‌ user Exposure Score
Keywords:
Human mobility, Privacy, Exposure‌
Scientific Description:

Human mobility data can support a wide range of applications,‌ from urban planning and‌ transportation optimization to epidemiological‌‌ modeling. However, the very same data that enables‌ such analyses often exposes‌ individuals to privacy risks‌‌ due to the strong uniqueness of their mobility‌ patterns. Current measures of‌ user exposure rely either‌‌ on simplistic heuristics—such as the number of unique‌ location sequences—or on specific‌ attack models, thereby limiting‌‌ their generalizability and interpretability.

To address these limitations,‌ we developed a customized‌ framework called MoBES (Mobility‌‌ Behavioral Exposure Score), which provides a flexible and‌ interpretable method to quantify‌ user exposure based on‌‌ behavioral mobility signatures. Rather than relying solely on‌ raw trajectories, MoBES extracts‌ a comprehensive set of‌‌ metrics from each user’s mobility traces.

Our implementation‌ computes user exposure by‌ measuring their distance from‌‌ behavioral neighbors in this feature space, independently of‌ any specific attack model.‌ This allows MoBES to‌‌ detect subtle but critical privacy differences in user‌ behavior that traditional measures‌ fail to capture. The‌‌ framework is modular, scalable, and applicable to large‌ datasets. It also includes‌ routines for comparative evaluation‌‌ against baseline measures, as well as visualization tools‌ to explore the distribution‌ of exposure scores across‌‌ the user population.
Functional Description:

Human mobility data‌ can support a wide‌ range of applications, from‌‌ urban planning and transportation optimization to epidemiological modeling.‌ However, the very same‌ data that enables such‌‌ analyses often exposes individuals to privacy risks due‌ to the strong uniqueness‌ of their mobility patterns.‌‌ Current measures of user exposure rely either on‌ simplistic heuristics—such as the‌ number of unique location‌‌ sequences—or on specific attack models, thereby limiting their‌ generalizability and interpretability.

To‌ address these limitations, we‌‌ developed a customized framework called MoBES (Mobility Behavioral‌ Exposure Score), which provides‌ a flexible and interpretable‌‌ method to quantify user exposure based on behavioral‌ mobility signatures. Rather than‌ relying solely on raw‌‌ trajectories, MoBES extracts a comprehensive set of metrics‌ from each user’s mobility‌ traces.

Our implementation computes‌‌ user exposure by measuring their distance from behavioral‌ neighbors in this feature‌ space, independently of any‌‌ specific attack model. This allows MoBES to detect‌ subtle but critical privacy‌ differences in user behavior‌‌ that traditional measures fail to capture. The framework‌ is modular, scalable, and‌ applicable to large datasets.‌‌ It also includes routines for comparative evaluation against‌ baseline measures, as well‌ as visualization tools to‌‌ explore the distribution of exposure scores across the‌ user population.
Release Contributions:‌
- 1st version of‌‌ the tool - An open-source deposit was requested,‌ and we are waiting‌ for signatures.
News of‌‌ the Year:
- 1st version of the tool‌ - An open-source deposit‌ was requested, and we‌‌ are waiting for signatures.
Publication:
hal-05057826
Contact:
Aline‌ Carneiro Viana
Participants:
Aline‌ Carneiro Viana, Lucas Felix‌‌ Da Silva, Jussara Marques Almeida, Nadjib Achir, Anne‌ Josiane Kouam Djuigne

7.1.12‌ ZenPlus

Name:
ZenPlus
Keywords:‌‌
Mobile networks, Mobility, Traffic data, Data fusion, Probabilistic‌ modeling, Prediction
Scientific Description:‌

ZenPlus provides a methodological‌‌ framework for jointly modeling‌ mobility and traffic behaviors of mobile network users‌ using fine-grained data such as eXtended Data Records‌ (XDRs).

The approach relies on a discrete abstraction‌ of user trajectories, where mobility and traffic time‌ series are transformed into sequences of interpretable behavioral‌ states. This representation captures temporal dynamics and inter-individual‌ heterogeneity while enabling generalization across geographical contexts.

Based‌ on this abstraction, ZenPlus implements a probabilistic likelihood‌ model that quantifies the compatibility between mobility and‌ traffic sequences. This model supports cross-modality matching, cross-domain‌ prediction (mobility to traffic and vice versa), and‌ data reconstruction.

The framework is both descriptive and‌ generative: it identifies structural dependencies between mobility and‌ network usage while enabling the generation of realistic‌ synthetic data. It is also transferable across regions,‌ as demonstrated on large-scale datasets.

Such an approach‌ enables applications including anomaly detection, quality-of-experience (QoE) optimization,‌ and realistic mobile network simulation.
Functional Description:

ZenPlus‌ is a Python library designed to analyze mobile‌ user behavior by combining two types of data:‌ mobility (movement) and network traffic (data usage). The‌ software converts raw data into simplified representations and‌ uses probabilistic models to capture the relationship between‌ where users are and how they consume data.‌

It enables users to: - understand the relationship‌ between mobility and data usage - match or‌ reconstruct datasets coming from separate sources - predict‌ behavior in one domain (mobility or traffic) from‌ the other - generate realistic synthetic data for‌ simulation or analysis

ZenPlus is intended for research‌ and advanced mobile network analysis, with applications such‌ as anomaly detection and quality-of-experience optimization.
URL:
https://gitlab.inria.fr/simbox-fraud-mitigation/legitimate-mobile-datasets-generation/zenplus_lib‌
Publication:
hal-05248595
Contact:
Anne Josiane Kouam Djuigne

7.1.13‌ MITIK-HAND

Name:
MITIK's data handling tool
Keywords:
Wi-Fi,‌ Trajectory Generation, Mobility
Functional Description:
MITIK-HAND comprises two‌ tools: 1. The first tool performs MAC association‌ of randomized MAC addresses used by the same‌ device from probe-requests. This tool models the frame‌ association to resolve MAC conflicts in small intervals.‌ It uses time and frame content-based signatures to‌ resolve and associate MACs inside a conflict. Finally,‌ a logistic regression-based algorithm using the obtained signatures‌ is proposed to associate devices with similar signatures.‌ 2. The second tool reconstructs a mobile terminal's‌ trajectory by introducing the concept of bounded trajectory.‌ It leverages the signal strength of users' public‌ WiFi probe requests collected from measurements of multiple‌ deployed or sniffers. Characterize and approximate the error‌ in the radial distances between the device and‌ the sniffer. Leverage the error characterization and approximated‌ radial distances to estimate the bounds associated with‌ a device's position. Finally, considering the spatiotemporal bounds‌ of device positions over time, it infers the‌ user's bounded trajectory.
Release Contributions:
Software and document‌ improvement.
News of the Year:
Software and document‌ improvement.
URL:
https://gitlab.inria.fr/fmorlano/mitik_data_handling
Contact:
Fernando Molano Ortiz
Participants:‌
Fernando Molano Ortiz, Aline Carneiro Viana, Nadjib Achir,‌ Abhishek Mishra, Guillaume Farhi Rivasseau

7.1.14 HEXPOSE

Name:‌
Hyperbox-based Exposure in a Mobility Behavioral Space
Keywords:‌
Mobility, Privacy, Exposure
Scientific Description:

The growing availability of large-scale human mobility‌ data supports a wide‌ range of applications, from‌‌ urban planning and transport optimization to climate action‌ and public health monitoring.‌ However, these benefits come‌‌ with significant privacy risks, as individual mobility patterns‌ are highly distinctive and‌ can enable re-identification. While‌‌ early work focused on spatio-temporal uniqueness, recent studies‌ have shown that behavioral‌ mobility patterns—captured through metrics‌‌ derived from trajectories—can also expose users to privacy‌ threats.

Existing techniques for‌ quantifying this behavioral exposure‌‌ often rely on computationally expensive sub-trajectory comparisons, make‌ strong assumptions about the‌ adversary’s knowledge, or underestimate‌‌ exposure risks by relying on overly simplified aggregate‌ statistics. To address these‌ limitations, we developed a‌‌ tailored methodological framework called Hyberbox (Hyperbox-based Exposure in‌ a Mobility Behavioral Space),‌ which offers a scalable‌‌ and interpretable method for quantifying user exposure based‌ on their behavioral mobility‌ signatures.

Instead of comparing‌‌ users via exhaustive pairwise distance calculations or attack-driven‌ models, Hyberbox projects each‌ user into a multidimensional‌‌ behavioral space and defines user-specific hyperboxes, based on‌ configurable variation thresholds for‌ different behavioral metrics. Users‌‌ whose behavioral profiles fall outside the hyperboxes of‌ all other users are‌ flagged as exposed, reflecting‌‌ a high degree of behavioral distinctiveness.
Functional Description:‌

The growing availability of‌ large-scale human mobility data‌‌ supports a wide range of applications, from urban‌ planning and transport optimization‌ to climate action and‌‌ public health monitoring. However, these benefits come with‌ significant privacy risks, as‌ individual mobility patterns are‌‌ highly distinctive and can enable re-identification. While early‌ work focused on spatio-temporal‌ uniqueness, recent studies have‌‌ shown that behavioral mobility patterns—captured through metrics derived‌ from trajectories—can also expose‌ users to privacy threats.‌‌

Existing techniques for quantifying this behavioral exposure often‌ rely on computationally expensive‌ sub-trajectory comparisons, make strong‌‌ assumptions about the adversary’s knowledge, or underestimate exposure‌ risks by relying on‌ overly simplified aggregate statistics.‌‌ To address these limitations, we developed a tailored‌ methodological framework called Hyberbox‌ (Hyperbox-based Exposure in a‌‌ Mobility Behavioral Space), which offers a scalable and‌ interpretable method for quantifying‌ user exposure based on‌‌ their behavioral mobility signatures.

Instead of comparing users‌ via exhaustive pairwise distance‌ calculations or attack-driven models,‌‌ Hyberbox projects each user into a multidimensional behavioral‌ space and defines user-specific‌ hyperboxes, based on configurable‌‌ variation thresholds for different behavioral metrics. Users whose‌ behavioral profiles fall outside‌ the hyperboxes of all‌‌ other users are flagged as exposed, reflecting a‌ high degree of behavioral‌ distinctiveness.
Release Contributions:
-‌‌ 1st version of the tool - An open-source‌ deposit was requested, and‌ we are waiting for‌‌ signatures.
News of the Year:
- 1st version‌ of the tool -‌ An open-source deposit was‌‌ requested, and we are waiting for signatures.
Publication:‌
hal-05057818
Contact:
Aline Carneiro‌ Viana
Participants:
Lucas Felix‌‌ Da Silva, Aline Carneiro Viana, Jussara Marques Almeida,‌ Nadjib Achir, Anne Josiane‌ Kouam Djuigne

7.2 New‌‌ platforms

Open Experimental IoT Platforms

Participants: Cedric Adjih‌, Francois-Xavier Molina,‌ Alexandre Abadie, Koen‌‌ Zandberg, Emmanuel Baccelli‌, Chetanveer Gobin, Fernando Molano, Mehdi‌ Debbah.

One necessity for research in the‌ domain of IoT is to establish and improve‌ IoT hardware platforms and testbeds that integrate representative‌ scenarios (such as Smart Energy, Home Automation, etc.)‌ and follow the evolution of technology, including radio‌ technologies and associated experimentation tools. For that, the‌ TRiBE team builds upon the evolutions of the‌ FIT IoT-LAB federated testbeds towards SLICES-FR (through our‌ 5G-mMTC-lab and NGC-AIoT platforms), which the team has‌ contributed to designing and deploying. We plans to‌ further develop FIT IoT-LAB with Edge AI and‌ more heterogeneous, up-to-date IoT hardware, and‌ radios to provide a usable and realistic experimentation‌ environment.

On the software side, IoT hardware available‌ so far has made it uneasy for developers‌ to build apps that run across heterogeneous hardware‌ platforms. For instance, Linux does not scale down‌ to small, energy- constrained devices, while microcontroller-based OS‌ alternatives were so far rudimentary and yield a‌ steep learning curve and lengthy development life-cycles because‌ they do not support standard programming and debugging‌ tools. As a result, another necessity for research‌ in this domain is to allow the emergence‌ of it more powerful, unifying IoT software platforms,‌ to bridge this gap. For that, we plan‌ to continue building upon RIOT, a new open‌ source software platform that provides a portable, Linux-like‌ API for heterogeneous IoT hardware. We plan to‌ continue to develop the systems and network stacks‌ aspects of RIOT, within the open source‌ developer community currently emerging around RIOT, amd also‌ provide an universal platform that can also be‌ used both (i) in the context of research‌ and/or teaching, as well as (ii) in industrial‌ contexts.

7.3 Open data

SigN Dataset: SIMBox Activity‌ Detection Through Cellular Latency Anomalies

Contributors:
Anne Josiane‌ Kouam, Aline Carneiro Viana, Philippe Martins, Cédric Adjih,‌ Alain Tchana.
Description:
This dataset contains latency and‌ cellular signaling measurements collected during controlled SIMBox detection‌ experiments on LTE networks. Data was gathered using‌ dedicated smartphones and SIMBox devices in both indoor‌ and outdoor environments under diverse signal conditions. The‌ dataset includes authentication latency measurements under multiple algorithms,‌ static attenuation tests, and IP-level interaction captures. No‌ user-identifiable information is present, as the experiments did‌ not involve real users.
Dataset PID (DOI,...):
10.57745/JBBBGO‌.
Project link:
Recherche Data Gouv.
Publications:‌
A. J. Kouam et al., SigN: SIMBox‌ Activity Detection Through Latency Anomalies at the Cellular‌ Edge, ACM AsiaCCS 2025.
Contact:
Anne Josiane‌ Kouam (anne-josiane.kouam@inria.fr).
Release contributions:
Dataset released by the‌ TRiBE team (version 2).

FraudZen Dataset: Realistic Ground‌ Truth CDRs of Bypass Fraud Techniques

Contributors:
Anne‌ Josiane Kouam, Aline Carneiro Viana, Alain Tchana.
Description:‌
This dataset contains synthetic call-detail records generated using‌ the open-source FraudZen simulator. It models legitimate mobile‌ subscribers and SIMBox fraudsters under multiple threat strategies,‌ including mobility-, traffic-, and social-based fraud scenarios. The‌ dataset spans different fraud prevalence levels and numbers of fraudulent SIM cards,‌ providing a realistic benchmark‌ for evaluating fraud detection‌‌ methods in cellular networks.
Dataset PID (DOI,...):
10.57745/TAV6PQ‌.
Project link:
Recherche‌ Data Gouv.
Publications:‌‌
A. J. Kouam et al., Battle of‌ Wits: To What Extent‌ Can Fraudsters Disguise Their‌‌ Tracks in International Bypass Fraud?, ACM AsiaCCS‌ 2024.
Contact:
Anne Josiane‌ Kouam (anne-josiane.kouam@inria.fr).
Release contributions:‌‌
Dataset released by the TRiBE team (version 2).‌

MITIK Dataset of pseudo-anonymized‌ WiFi management frames from‌‌ a parking area

Contributors:
Fernando Molano Ortiz; Abhishek‌ Kumar Mishra; Aline Carneiro‌ Viana; Nadjib Achir.
Description:‌‌
This dataset (MITIK-PARKING) is part of the ANR‌ MITIK project. It contains‌ WiFi management frames—specifically probe‌‌ requests, probe responses, and beacons—captured from two test‌ devices with non-randomized MAC‌ addresses. These frames were‌‌ recorded in parallel by multiple super-sniffers operating on‌ channel 1 of the‌ 2.4 GHz WiFi band‌‌ recorded these frames in parallel. Data collection was‌ performed synchronously using the‌ MITIK-MGMT (https://gitlab.inria.fr/mitik/measurement-management/mitik-mgmt)toolkit, and the‌‌ frames are stored in .pcap format. For more‌ details, please refer to‌ the README.md file.
Dataset‌‌ PID (DOI,...):
.
Project link:
Recherche Data Gouv‌.
Publications:
Fernando Dias‌ de Mello Silva, Abhishek‌‌ Kumar Mishra, Aline Carneiro Viana, Nadjib Achir, Anne‌ Fladenmuller, et al.. Performance‌ Analysis of a Privacy-Preserving‌‌ Frame Sniffer on a Raspberry Pi. CSNet 2022‌ - 6th Cyber Security‌ in Networking Conference, Oct‌‌ 2022, Rio de Janeiro, Brazil. ⟨10.1109/CSNet56116.2022.9955615⟩. ⟨hal-03906600⟩ doi:‌ 10.1109/CSNet56116.2022.9955615.
Contact:
Nadjib Achir‌ (nadjib.achir@inria.fr).
Release contributions:
Dataset‌‌ released by the TRiBE team (version 2).

MITIK‌ Dataset of WiFi pseudo-anonymised‌ public management frames captured‌‌ at La Rochelle University

Contributors:
Fernando Dias de‌ M. Silva; Fernando Molano‌ Ortiz; Abhishek Kumar Mishra;‌‌ Antoine Huchet; Mohammad Imran Syed; Luís Henrique M.‌ K. Costa; Anne Fladenmuller;‌ Aline Carneiro Viana; Yacine‌‌ Ghamri-Doudane; Nadjib Achir.
Description:
This dataset consists of‌ WiFi management frames (probe-requests,‌ probe-responses, and beacons) captured‌‌ during a passive in-field measurement campaign using the‌ MITIK-SENS tool version 1‌ at La Rochelle University.‌‌ Data collection was performed by five supersniffers (each‌ with five sniffers) operating‌ on WiFi channel 1‌‌ (2.4 GHz) across two scenarios, with four 60-minute‌ experiments conducted. Pseudo-anonymization of‌ MAC addresses is applied‌‌ to ensure GDPR compliance. Captured frames are stored‌ as .pcap files. For‌ more details, please refer‌‌ to the README.md file.
Dataset PID (DOI,...):
.‌
Project link:
Recherche Data‌ Gouv.
Publications:
Fernando‌‌ Dias de Mello Silva, Abhishek Kumar Mishra, Aline‌ Carneiro Viana, Nadjib Achir,‌ Anne Fladenmuller, et al..‌‌ Performance Analysis of a Privacy-Preserving Frame Sniffer on‌ a Raspberry Pi. CSNet‌ 2022 - 6th Cyber‌‌ Security in Networking Conference, Oct 2022, Rio de‌ Janeiro, Brazil. ⟨10.1109/CSNet56116.2022.9955615⟩. ⟨hal-03906600⟩‌ doi: 10.1109/CSNet56116.2022.9955615.
Contact:
Nadjib‌‌ Achir (nadjib.achir@inria.fr).
Release contributions:
Dataset released by the‌ TRiBE team (version 2).‌

8 New results

Activities‌‌ and related result hereafter described are classified according‌ to the three research‌ axes of the team.‌‌

8.1 [Axis 1]: Optimized communication protocols

8.1.1 Modern‌ Random Access: Irregular Repetition‌ Slotted Aloha (IRSA)

Participants:‌‌ Andrei-Valentin Stirbu, Cédric‌ Adjih, Paul Mühlethaler [Inria, EVA], Chung‌ Shue Chen [Nokia Bell Labs], Pengwenlong Gu‌, Saeed Alsabbagh [Université Paris-Saclay - UVSQ, France‌ & Laboratoire DAVID, France], Nadjib Aitsaadi [UVSQ‌ Paris-Saclay & DAVIDLab, France], Amine Adouane [Benyoucef‌ Benkhedda University, Algeria].

Wireless communications play an‌ important part in the systems of the Internet‌ of Things (IoT). Recently, there has been a‌ trend towards long-range communications systems for the IoT,‌ including cellular networks. For many use cases, such‌ as massive machine-type communications (mMTC), performance can be‌ gained by moving away from the classical model‌ of connection establishment and adopting random access methods.‌ Associated with physical layer techniques such as Successive‌ Interference Cancellation (SIC), or Non-Orthogonal Multiple Access (NOMA),‌ the performance of random access can be dramatically‌ improved, giving rise to novel random access protocol‌ designs.

In this line of work, we are‌ studying a modern method of random access for‌ packet networks, named “Irregular Repetition Slotted Aloha (IRSA)”,‌ that had been recently proposed: it is based‌ on repeating transmitted packets and on the use‌ of successive interference cancellation at the receiver. In‌ classical idealized settings of slotted random access protocols‌ (where slotted ALOHA achieves 1/e), it has been‌ shown that IRSA could asymptotically achieve the maximal‌ throughput of 1 packet per slot.

8.1.2 An‌ Introduction to Modern Random Access Protocols for IoT‌ Communications (book chapter)

Participants: Iman Hmedoush, Jia‌ Cao, Cédric Adjih, Sanjeev Sharma [IIT‌ (BHU) Varanasi, India], Kuntal Deka [IIT Guwahati,‌ India].

In this book chapter, we‌ present a perspective on a category of grant-free‌ communication methods, specifically those based on random access.‌ They are well-suited for scenarios involving a large‌ number of devices transmitting sporadically with short messages,‌ commonly encountered in Internet-of-Things (IoT) communication scenarios. These‌ methods, often referred to as modern random access,‌ have recently emerged and generated significant interest. We‌ discuss, in an introductory and tutorial manner, the‌ basic methods for evaluating these communications as introduced‌ in the literature. Additionally, we provide insights into‌ their performance.

8.1.3 Optimization of Irregular Repetition Slotted‌ ALOHA with Imperfect SIC in 5G CIoT

Participants:‌ Saeed Alsabbagh [Université Paris-Saclay - UVSQ, France &‌ Laboratoire DAVID, France], Cédric Adjih, Amine‌ Adouane [Benyoucef Benkhedda University, Algeria], Nadjib Aitsaadi‌ [UVSQ Paris-Saclay & DAVIDLab, France].

Irregular Repetition‌ Slotted ALOHA (IRSA) is an effective grant-free random‌ access scheme that is well-suited for managing the‌ sporadic nature of IoT traffic, particularly in dense‌ environments prone to collisions. In this paper, we‌ evaluate the performance of IRSA under realistic conditions‌ involving imperfect successive interference cancellation (SIC) and non-ideal‌ physical layer environments. Specifically, we investigate the impact‌ of various channel conditions and physical layer impairments‌ on IRSA's performance. Previous studies on IRSA often‌ assume ideal physical layer conditions or use simplified‌ models for SIC errors, which fail to fully‌ capture practical implementation complexities. To address this gap, we propose integration of‌ practical factors, such as‌ channel estimation imperfections, into‌‌ our model of SIC failures using detailed baseband‌ simulations. Based on that,‌ we employ density evolution‌‌ analysis to evaluate system throughput and optimize the‌ degree distributions to enhance‌ IRSA performance in the‌‌ presence of imperfect SIC. Our results focusing on‌ 5G CIoT demonstrate that‌ optimizing IRSA parameters, while‌‌ accounting for SIC errors, can significantly improve system‌ performance, resulting in notable‌ throughput gains.

This article‌‌ 20 was presented at ICC 2025.

8.1.4 IRSA‌ Under Capture Effect and‌ Imperfect SIC: a de‌‌ Analysis for Future Cellular IoT

Participants: Saeed Alsabbagh‌ [Université Paris-Saclay - UVSQ,‌ France & Laboratoire DAVID,‌‌ France], Cédric Adjih, Amine Adouane [Benyoucef‌ Benkhedda University, Algeria],‌ Nadjib Aitsaadi [UVSQ Paris-Saclay‌‌ & DAVIDLab, France].

Irregular Repetition Slotted ALOHA‌ (IRSA) is a leading‌ candidate for random access‌‌ and grant-free communication in future Cellular IoT (CIoT)‌ networks, including those envisioned‌ for 6G and beyond.‌‌ Classical analyses of IRSA typically assume ideal conditions;‌ however, real deployments are‌ subject to practical impairments.‌‌ In particular, the capture effect enables packets to‌ be decoded despite collisions‌ when their signal-to-interference ratios‌‌ exceed certain thresholds, and imperfect successive interference cancellation‌ (SIC), due to channel‌ estimation errors, further complicates‌‌ decoding dynamics.

In this paper, we are the‌ first to develop a‌ unified analytical framework that‌‌ incorporates both phenomena into the IRSA design. Using‌ a threshold-based capture model‌ and a detailed residual‌‌ interference analysis, we apply density evolution to derive‌ asymptotic throughput bounds. Our‌ results show that by‌‌ optimizing the user degree distribution, IRSA can significantly‌ mitigate performance loss under‌ non-ideal SIC conditions. Extensive‌‌ simulations validate our theoretical findings, revealing that performance‌ improvements are attainable even‌ in high-density CIoT scenarios.‌‌

This article 19 was presented at PIMRC 2025.‌

8.1.5 FIT-IRSA: Feedback-Integrated Two-Phase‌ IRSA with Deep Reinforcement‌‌ Learning

Participants: Andrei-Valentin Stirbu, Cédric Adjih.‌

Efficient random access can‌ be used in scenarios‌‌ with a massive number of IoT devices. Among‌ modern random access protocols,‌ Irregular Repetition Slotted ALOHA‌‌ (IRSA) offers excellent asymptotic performance (for large frame‌ sizes), but its finite-frame‌ efficiency is lower and‌‌ difficult to optimize analytically. In this work, we‌ introduce limited mid-frame feedback‌ to better coordinate users‌‌ and improve performance: Feedback-Integrated Two-phase IRSA (FIT-IRSA). We‌ formulate IRSA with feedback‌ as a deep reinforcement‌‌ learning (DRL) problem. Using policy gradient methods, we‌ learn transmission strategies that‌ improve throughput under varying‌‌ loads, as demonstrated in our simulation results. This‌ provides a practical alternative‌ to classical density-evolution–based optimization,‌‌ which applies mainly to large frames.

A preliminary‌ version work was presented‌ at the junior conference‌‌ JWOC 2025, and a final version 42‌ was submitted and presented‌ at PEMWN 2025.‌‌

8.1.6 Terahertz Communication: State-of-the-Art and Future Directions

Participants:‌ Sanjeev Sharma [IIT (BHU)‌ Varanasi], Praveen K.‌‌ Singya [IIITM Gwalior], Kuntal Deka [IIT Guwahati]‌, Cédric Adjih,‌ Mohit Sharma [Technology Innovation‌‌ Institute, Abu Dhabi, UAE]‌.

Terahertz (THz) communication is a cutting-edge technology‌ poised to address the increasing demand for ultra-high‌ data rates in next-generation wireless networks. THz frequencies‌ offer a vast spectrum supporting ultra-massive connectivity for‌ applications like high-definition video streaming, virtual reality, and‌ advanced Internet-of-Things (IoT) devices. Additionally, THz communication can‌ enable new technologies such as high-resolution imaging and‌ sensing. This paper presents a comprehensive survey on‌ THz communication’s advancements, challenges, and potential applications. We‌ provide an overview of the fundamental concepts, including‌ THz band characteristics, transmission mechanisms, and channel models.‌ Additionally, we discuss the current state-of-the-art in THz‌ communication technology, covering aspects such as modulation techniques,‌ radio frequency (RF) front-end design, and signal processing‌ algorithms. Furthermore, we analyze the challenges and limitations‌ associated with THz communication, such as propagation losses,‌ atmospheric absorption, beam split, and hardware constraints. This‌ paper aims to provide researchers and practitioners with‌ a comprehensive understanding of THz communication design and‌ analysis and inspire further advancements in this rapidly‌ evolving field.

This 18 is published in the‌ IEEE Open Journal of the Communication Society.

8.2‌ [Axis 1] Tailored embedded software platforms and Unified‌ low-end IoT technology

8.2.1 Embedded Rust Operating System‌ for Networked Sensors and Multi-Core Microcontrollers

Participants: Elena‌ Frank, Koen Zandberg, Emmanuel Baccelli,‌ Romain Fouquet, Kaspar Schleiser, Christian Amsüss‌.

Large swaths of low-level system software building‌ blocks originally implemented in C/C++ are currently being‌ swapped for equivalent rewrites in Rust, a relatively‌ more secure and dependable programming language. So far,‌ however, no embedded OS in Rust supports multicore‌ preemptive scheduling on microcontrollers. In this paper, we‌ thus fill this gap with a new operating‌ system: Ariel OS. We describe its design, we‌ provide the source code of its implementation, and‌ we perform micro-benchmarks on the main 32-bit microcontroller‌ architectures: ARM Cortex-M, RISC-V and Espressif Xtensa. We‌ show how our scheduler takes advantage of several‌ cores, while incurring only small overhead on single-core‌ hardware. As such, Ariel OS provides a convenient‌ embedded software platform for small networked devices, for‌ both research and industry practitioners.

This work is‌ detailed in 29, published in the International‌ Conference on Distributed Computing in Smart Systems and‌ the Internet of Things (DCOSS-IoT), in May 2025.‌

8.2.2 SLAs for Shared Responsibility in Multi-Tenant Microcontrollers‌

Participants: Bastien Buil, Chrystel Gaber, Sylvain‌ Plessis, Emmanuel Baccelli, Samia Bouzefrane.‌

Lightweight software containerization solutions execute multiple payloads from‌ several mutually distrusting stakeholders on a resource-constrained microcontroller.‌ This paradigm shifts the accountability model from a‌ single-accountable-actor model where there is only one integrator‌ responsible for the entire monolithic code to a‌ multiple-accountable-actor model where multiple stakeholders share responsibilities. This‌ paper explores this model through three dimensions: responsibility‌ distribution, fulfillment of Cloud commitments, and commitment verification‌ mechanisms along with stakeholders' access to them.

This‌ work is detailed in 22, published in‌ the International Conference on Network and Service Management (CNSM), in October 2025.‌

8.3 [Axis 1] Low-footprint‌ cybersecurity mechanisms

8.3.1 Secure‌‌ Low-Power Software Continuous Deployment

Participants: Frédéric Fort,‌ Koen Zandberg, Emmanuel‌ Baccelli, Hugo Forraz‌‌, Gilles Grimaud.

Continuous deployment (CD) is‌ often a bottleneck for‌ software running on microcontrollers‌‌ (MCUs). CD remains a challenge to this day‌ because software updates for‌ MCUs lack convenient and‌‌ secure partial updates mechanisms. Updates thus remain predominantly‌ monolithic (firmware updates). To‌ bridge this gap, we‌‌ design PURR, a solution combining a formally verified‌ memory partition mechanism (PIP),‌ and tiny software virtualization‌‌ (rBPF) which we integrate in a common operating‌ system (RIOT). PURR enable‌ secure software enclaves on‌‌ microcontrollers with a memory protection unit (MPU), which‌ can be updated securely‌ over the network, and‌‌ which can eXecute in Place (XiP) a small‌ virtual machine following the‌ eBPF instruction set architecture.‌‌ We publish an open source implementation of PURR‌ and we provide results‌ of benchmarks on a‌‌ popular Arm Cortex-M microcontroller. Our experiments show that‌ the additional mechanisms guaranteeing‌ PURR enclaves' (formally verified)‌‌ security enable substantial rBPF execution speed improvements, while‌ incurring a modest memory‌ footprint overhead.

This work‌‌ is detailed in 28, published in the‌ International Conference on Distributed‌ Computing in Smart Systems‌‌ and the Internet of Things (DCOSS-IoT), in May‌ 2025.

8.3.2 Standardization of‌ Secure Software Updates for‌‌ Low-power IoT Devices

Participants: Koen Zandberg, Emmanuel‌ Baccelli.

TRiBE co-authors‌ the new IETF standard‌‌ (work-in-progress) providing low-end IoT devices with secure software‌ updates. The Internet Draft‌ draft-ietf-suit-manifest-34 specifies a Concise‌‌ Binary Object Representation (CBOR)-based Serialization Format for the‌ Software Updates for Internet‌ of Things (SUIT) Manifest.‌‌ This specification describes the format of a manifest.‌ A manifest is a‌ bundle of metadata about‌‌ the firmware for an IoT device, where to‌ find the firmware, the‌ devices to which it‌‌ applies, and cryptographic information protecting the manifest. Firmware‌ updates and secure boot‌ both tend to use‌‌ sequences of common operations, so the manifest encodes‌ those sequences of operations,‌ rather than declaring the‌‌ metadata. The manifest also serves as a building‌ block for secure boot.‌

This work was published‌‌ in the IETF Internet Draft draft-ietf-suit-manifest-34, in‌ May 2025.

8.4 [Axis‌ 1] Edge AI and‌‌ Tiny Machine Learning

8.4.1 TinyML as a Service‌ on Multi-Tenant Microcontrollers

Participants:‌ Bastien Buil, Chrystel‌‌ Gaber, Emmanuel Baccelli, Samia Bouzefrane.‌

Tiny Machine Learning (TinyML)‌ allows the execution of‌‌ small machine learning models on low-power devices like‌ microcontrollers. TinyML-as-a-Service (TinyMLaaS) is‌ an architecture to make‌‌ the usage of TinyML models easier by having‌ a platform that optimizes‌ and compiles machine learning‌‌ models according to the constraints of target devices,‌ and then deploys the‌ model code on microcontrollers.‌‌ Within the Cloud-to-IoT continuum, both TinyML and multi-tenant‌ microcontrollers focus on empowering‌ microcontrollers and enabling on-device‌‌ computing. Multi-tenant microcontrollers are designed to securely execute‌ codes from mutually distrusting‌ actors through the usage‌‌ of lightweight software containerization‌ solutions, like WebAssembly. In this paper, we propose‌ to integrate TinyMLaaS with multi-tenant microcontrollers by using‌ WebAssembly-based containerization, and we implement a proof-of-concept of‌ the TinyMLaaS architecture based on We-bAssembly Micro Runtime‌ (WAMR) and RIOT-ML. In the second part of‌ the paper, to improve the usage of containerized‌ TinyML on microcontrollers, we propose CS4WAMR, a framework‌ to enhance WAMR usage by enabling running simultaneously‌ multiple instances of WAMR to allow better permission‌ and memory consumption control.

This work is detailed‌ in 21, published in the ACM International‌ Conference on Embedded Wireless Systems and Networks (EWSN),‌ in September 2025.

8.4.2 RAM Footprint Optimizations for‌ TinyML Inference with Convolutional Neural Networks

Participants: Zhaolan‌ Huang, Emmanuel Baccelli.

AI spans from‌ large language models to tiny models running on‌ microcontrollers (MCUs). Extremely memory-efficient model architectures are decisive‌ to fit within an MCU's tiny memory budget‌ e.g., 128kB of RAM. However, inference latency must‌ remain small to fit real-time constraints. An approach‌ to tackle this is patch-based fusion, which aims‌ to optimize data flows across neural network layers.‌ In this paper, we introduce msf-CNN, a novel‌ technique that efficiently finds optimal fusion settings for‌ convolutional neural networks (CNNs) by walking through the‌ fusion solution space represented as a directed acyclic‌ graph. Compared to previous work on CNN fusion‌ for MCUs, msf-CNN identifies a wider set of‌ solutions. We published an implementation of msf-CNN running‌ on various microcontrollers (ARM Cortex-M, RISC-V, ESP32). We‌ show that msf-CNN can achieve inference using 50%‌ less RAM compared to the prior art (MCUNetV2‌ and StreamNet). We thus demonstrate how msf-CNN offers‌ additional flexibility for system designers.

This work is‌ detailed in 30, published in the Annual‌ Conference on Neural Information Processing Systems (NeurIPS), in‌ December 2025.

8.5 [Axis 2] Extracting high-end IoT‌ footprints in data.

8.5.1 MITIK Toolkit: A Privacy-Compliant‌ Passive Collection of WiFi Probe Request Datasets [Axis‌ 2]

Participants: Fernando Molano Ortiz, Guillaume Farhi-Rivasseau‌, Nadjib Achir, Aline Carneiro Viana.‌

The ubiquity of WiFi-connected devices broadcasting unencrypted management‌ frames enables identifying nearby devices, which can be‌ beneficial for societal applications while raising significant privacy‌ concerns. This demo paper introduces a unified toolkit,‌ Mitik, for capturing, analyzing, and interpreting non-intrusive passive‌ measurements of WiFi traces. The toolkit addresses several‌ challenges, including the configuration of sniffers for synchronized‌ data capture, privacy protection at the point of‌ collection, and the association of randomized MAC addresses‌ with individual smart-phones. By systematically tackling these challenges,‌ Mitik aims to advance our understanding of individual‌ mobility patterns and uncover plausible links between distinct‌ devices.

This work is related to the ANR‌ MITIK project (2020-2025) and was accepted to be‌ published as a demo at ACM SIGCOMM 2025‌ 33.

8.5.2 Towards assessing accessibility resulting from‌ integrating demand-responsive and conventional public transport with travel‌ time uncertainty

Participants: Mohamed Ourahou, Andrea Araldo‌ [Telecom SudParis-IPP, France], Louis Zigrand [PADAM Mobility, France], Aline Carneiro‌ Viana.

Transportation networks‌ provide people with the‌‌ means to reach opportunities (i.e., accessibility), e.g., schools‌ and workplaces [1, 2].‌ Conventional Public Transportation (CPT),‌‌ e.g., fixed-route buses and trains, is most effective‌ in densely populated areas,‌ where it offers extensive‌‌ and affordable coverage [3]. In low-density areas (e.g.,‌ suburbs), CPT provides insufficient‌ accessibility, as few opportunities‌‌ can be reached within reasonable travel times. Operators‌ limit frequency and coverage‌ to prevent high costs‌‌ per passenger [4], leaving such areas car-dependent, with‌ private vehicles often the‌ only viable mode [5].‌‌ Demand Responsive Transportation (DRT) offers a flexible alternative,‌ as routes and schedules‌ adapt to users’ requests.‌‌ With growing interest in integrating DRT with CPT,‌ it is essential to‌ evaluate its impact on‌‌ accessibility, particularly when DRT operates as stop-to-stop service,‌ both as a feeder‌ and as a standalone‌‌ mode. Yet, rigorous methods to assess accessibility under‌ these conditions remain lacking.‌ To address this gap,‌‌ we propose a method that accounts for travel‌ time uncertainty in DRT,‌ rather than relying on‌‌ deterministic or oversimplified assumptions.

This is the on-going‌ work of M. Ourahou's‌ PhD thesis. It was‌‌ published at the NetMob 2025 conference 41.‌

8.6 [Axis 2] User‌ pattern understanding at the‌‌ Internet edge.

8.6.1 Generating Random Hyperfractal Cities

Participants:‌ Philippe Jacquet, Bernard‌ Mans, Geoffrey Deperle‌‌.

In a previous work 68, we‌ provide analytic bounds on‌ the requirements in terms‌‌ of connectivity extension for vehicular networks served by‌ fixed Enhanced Mobile BroadBand‌ (eMBB) infrastructure, where both‌‌ vehicular networks and infrastructures are modeled using stochastic‌ and fractal geometry as‌ a model for urban‌‌ environments. The hyperfractal model can be used to‌ model cities with very‌ few parameters. Furthermore it‌‌ can be run as a generative models to‌ create an unbounded number‌ of imaginary cities for‌‌ AI training. This subject relates to the on-going‌ PhD thesis of Geoffrey‌ Deperle and the PEPR‌‌ MOBIDEC Mob Sci-Dat Factory project.

In this context,‌ the first published work‌ focuses on the challenge‌‌ of interactively modeling street networks. We extend the‌ simple fractal model, which‌ is particularly useful for‌‌ describing small cities or individual districts, by constructing‌ random cities based on‌ a tiling structure over‌‌ which hyperfractals are distributed. This approach enables the‌ connection of multiple hyperfractal‌ districts, providing a more‌‌ comprehensive urban representation. Furthermore, we demonstrate how this‌ decomposition can be used‌ to segment a city‌‌ into distinct districts through fractal analysis. Finally, we‌ present tools for the‌ numerical generation of random‌‌ cities following this model. This work was published‌ at the GSI 2025‌ conferences 23.

8.6.2‌‌ Predicting Mobility with Small Data and Physical Principles‌

Participants: Haron Calegari Fanticelli‌, Antonio Tadeu A.‌‌ Gomes, Aline Carneiro Viana.

The study‌ of human mobility is‌ fundamental due to its‌‌ impact on urban planning, epidemic spreading, population well-being,‌ and pollution mitigation, among‌ other applications. Despite significant‌‌ progress, key challenges remain,‌ notably the limited interpretability and generality of existing‌ models, as well as the strong imbalance in‌ data availability across regions. In many areas, scarce‌ or incomplete data prevent the effective use of‌ data-hungry mobility models.

To address these challenges, this‌ work proposes a novel approach that combines mathematical‌ models inspired by natural phenomena—typically expressed through differential‌ equations—with established machine learning techniques for mobility prediction.‌ This hybrid modeling strategy aims to inject domain‌ knowledge into data-driven methods, improving model interpretability while‌ reducing the reliance on large training datasets.

The‌ study focuses on aggregate mobility prediction, leveraging data‌ describing hourly flows of people between administrative regions‌ of Paris over a fourteen-day period. By modeling‌ visitation routines, the goal is to predict area-level‌ population density at given time instants. The central‌ research question is whether human visitation routines can‌ be accurately modeled using this combined mathematical and‌ machine learning framework, thereby enabling more interpretable and‌ data-efficient mobility prediction models with potential applications in‌ urban planning and epidemiology.

Haron C. Fanticelli is‌ a partial time PhD candidate. He has defended‌ the two required PhD follow-up examens and the‌ work is still on-going.

8.6.3 A Fine-Grained Analysis‌ of Individual Mobility and Traffic Dependencies

Participants: Anne‌ Josiane Kouam, Aline Carneiro Viana, Mariano‌ Beiro, Leo Ferres, Luca Pappalardo.‌

Understanding mobile-user behavior requires jointly modeling mobility and‌ mobile traffic, as data consumption is intrinsically shaped‌ by where, when, and how users move. Despite‌ this strong interdependence, most prior work treats mobility‌ and traffic in isolation, overlooking fine-grained behavioral dependencies‌ at the individual level. In this work, we‌ propose a framework that explicitly captures the interplay‌ between mobility and traffic behaviors using fine-grained mobile‌ datasets.

Leveraging week-long eXtended Data Records (XDRs), we‌ identify a compact set of interpretable features and‌ the mobility traits that drive traffic variations, enabling‌ a privacy-preserving user abstraction based on discrete mobility–traffic‌ states. We further introduce a probabilistic likelihood model‌ to assess the coherence of mobility–traffic pairings, supporting‌ cross-modality inference and the fusion of fragmented logs.‌ Experiments on large-scale datasets covering more than 1.3‌ million users show that the approach generalizes across‌ heterogeneous regions, paving the way for privacy-aware anomaly‌ detection, personalized QoE adaptation, and realistic network and‌ mobility simulation.

This work was presented at the‌ A-ranked ACM International Conference on Modeling, Analysis and‌ Simulation of Wireless and Mobile Systems (MSWiM) 2025‌, where it received the Best Paper Award‌ 32.

8.6.4 Assessing Usability and Reliability of‌ Anonymized Spatio-Temporal Data

Participants: Gaelle Yonga, Anne‌ Josiane Kouam, Aline Carneiro Viana, Auguste‌ Noumsi.

Spatio-temporal datasets, such as mobile network‌ traces and Charging Data Records (CDRs), are widely‌ reused to study human activity in domains including‌ urban planning, transportation, and infrastructure optimization. While anonymization‌ is essential to protect users’ privacy, it may‌ also introduce biases or distortions that compromise data‌ reliability and lead to misleading conclusions if not properly assessed beforehand.

In‌ this work, we highlight‌ the necessity of a‌‌ systematic pre-analysis characterization phase prior to reusing anonymized‌ spatio-temporal data. We propose‌ a generic and reusable‌‌ methodology to evaluate dataset usability along three complementary‌ dimensions: (i) global dataset‌ overview, (ii) traffic pattern‌‌ analysis, and (iii) mobility pattern analysis. This framework‌ enables researchers and practitioners‌ to identify structural biases,‌‌ assess behavioral realism, and determine application-specific limitations of‌ anonymized datasets.

We apply‌ the proposed methodology to‌‌ a large-scale anonymized CDR dataset from Shenzhen, revealing‌ critical insights regarding its‌ spatial representativeness, temporal consistency,‌‌ and mobility realism. Our results demonstrate that anonymization‌ and collection processes can‌ significantly affect downstream analyses,‌‌ and that such effects must be explicitly quantified‌ to ensure responsible and‌ meaningful data reuse. This‌‌ work contributes to more robust, privacy-aware, and context-sensitive‌ exploitation of spatio-temporal datasets.‌

This work was published‌‌ at CoReS 2025 36, and an extended‌ version was presented at‌ 7th International Workshop on‌‌ Urban Computing (UrbCom), co-located with IEEE DCOSS-IoT‌ 2025 37.

8.7‌ [Axis 2] Addressing end-users‌‌ privacy exposure and security concerns in networking data.‌

8.7.1 Behavior-based User Exposure‌ in Mobility Data

Participants:‌‌ Lucas Gabriel Da Silva Felix, Anne Josiane‌ Kouam Djuigne, Aline‌ Carneiro Viana, Nadjib‌‌ Achir, Jussara Almeida [Federal University of Minas‌ Gerais, Brazil].

Individual-level‌ mobility data drives major‌‌ economic value and application advances but entails significant‌ privacy risks due to‌ the high distinctiveness of‌‌ movement patterns, which are prone to re- identification.‌ The tension between analytics‌ benefits and privacy risk‌‌ poses a central requirement: developing flexible solutions to‌ measure and interpret user‌ exposure for targeted protection‌‌ guidance while support- ing mobility-driven services. Existing strategies‌ to quantify such user‌ exposure either focus only‌‌ on the sequences of places visited by each‌ user, as the widely‌ used uniqueness measure, or‌‌ are tied to specific attack models. In this‌ work, we explore a‌ complementary perspective on exposure‌‌ by focusing on persistent mobility behaviors such as‌ routines and mobility traits,‌ which shape how individuals‌‌ can be distinguished beyond the locations they visit.‌

We first designed MoBES,‌ a customizable measure of‌‌ user exposure in mobility data, leveraging multiple existing‌ metrics to build a‌ multi-dimensional space, which in‌‌ turn is used to capture each user's mobility‌ signature behavior. MoBES quantifies‌ user exposure based on‌‌ how distinct a user's signature is from her‌ neighbors in the defined‌ metric space. As such,‌‌ MoBES is designed to be a fundamental expression‌ of user behavior, and‌ not tied to any‌‌ specific attack model. We evaluate MoBES on a‌ real mobility dataset, showing‌ that it effectively captures‌‌ user exposure within the behavioral metric space. We‌ also compare MoBES with‌ the uniqueness measure, showing‌‌ that MoBES is able to uncover users who,‌ even though visiting the‌ same places as others‌‌ in the crowd, are still at risk of‌ exposure due to the‌ unicity of their mobility‌‌ behavior. This work was‌ published at the: IEEE MDM 2025 conference 27‌, SBRC Brazilian 2025 40, and at‌ the NetMob 2025 26 conferences. An extended version‌ is under a submission to EPJ Data Science.‌

However, MoBES relies on an aggregated score (averaging‌ distances across metrics and neighbors), which can mask‌ exposure driven by localized deviations and under-estimate a‌ user's true risk. To tackle this drawback, we‌ designed a new framework for measuring individual exposure‌ risk in mobility data. This new approache enables‌ both quantification and interpretation of users' exposure levels,‌ independently of the data collection process. This current‌ approach is under submission.

8.7.2 Sensor-based fingerprinting and‌ privacy risks in mobile systems

Participants: Carlos Sulbaran‌ Fandino, Anne Josiane Kouam, Konrad Rieck‌.

Modern mobile devices embed a wide range‌ of motion sensors that continuously capture fine-grained physical‌ signals. While primarily designed to support benign functionalities‌ such as activity recognition and gaming, these sensors‌ have also been shown to enable device fingerprinting‌ through subtle hardware-induced variations. Such fingerprinting poses a‌ significant privacy threat, as it allows persistent user‌ or device identification without relying on explicit identifiers,‌ permissions, or network-layer data.

In this work, we‌ provide the first comprehensive systematization of motion sensor-based‌ fingerprinting in mobile systems. We structure the fingerprinting‌ pipeline into distinct stages and identify the key‌ design choices, experimental assumptions, and evaluation metrics used‌ across the literature. Building on this analysis, we‌ introduce a unified and reproducible evaluation framework that‌ enables the systematic assessment of fingerprinting effectiveness under‌ realistic conditions, independently analyzing the impact of sensor‌ types, feature extraction strategies, classifiers, and acquisition settings.‌

Our experimental results demonstrate that motion sensor-based fingerprinting‌ remains highly effective across diverse scenarios and learning‌ models, while existing countermeasures provide limited protection and‌ often significantly degrade data utility. By releasing both‌ datasets and evaluation tools, this work supports reproducibility‌ and facilitates future research on this persistent privacy‌ threat. These findings highlight the urgent need for‌ stronger, utility-preserving defenses against sensor-based identification in mobile‌ platforms.

This work was published and presented at‌ the 27th International Conference on Modeling, Analysis and‌ Simulation of Wireless and Mobile Systems (ACM MSWiM‌ 2025) 25.

8.7.3 SigN: Empirical Insights and‌ Practical Solution for SIMBox Fraud Prevention at the‌ Cellular Edge

Participants: Anne Josiane Kouam, Aline‌ Carneiro Viana, Alain Tchana.

Cellular SIMBox‌ fraud bypasses international mobile calls and routes them‌ through the internet as local mobile calls in‌ the destination country, using VoIP GSM gateways equipped‌ with multiple SIM cards, also known as "SIMBox."‌ This fraud causes annual financial losses of up‌ to $3.11 billion, national security threats, and phone‌ conversation privacy breaches. Current approaches to mitigate SIMBoxx‌ fraud present open issues that affect their effectiveness.‌ They lack robustness against the constant refinement of‌ fraudsters' strategies or involve a certain implementation complexity‌ that hinders their widespread deployment in operator networks.‌

This work presents SigN, a new mitigation approach based on cellular signaling‌ data analysis. SigN is‌ the first-of-the-literature real-time prevention‌‌ methodology that is beyond fraudster-reach and largely deployable‌. SigN focuses on‌ the cellular signaling of‌‌ user devices during the network attachment, aiming‌ to block fraudulent SIMBox‌ devices before they can‌‌ connect to the network. Through extensive indoor and‌ outdoor experimentation, we empirically‌ show that fraudulent SIMBox‌‌ devices cause significant latency than legitimate devices during‌ the network attachment. Especially‌ in the authentication phase,‌‌ fraudulent SIMBox devices' minimum latency is 23 $\times‌$ higher than their legitimate‌ counterparts. We analyze such‌‌ latency overhead, showing it is fundamentally shaped by‌ factors beyond fraudsters' control,‌ i.e., LTE standards for‌‌ authentication and Internet-based communication related protocols and vagaries.‌ Therefore, we propose a‌ SIMBox fraud prevention approach‌‌ that adapts the standardized authentication procedure at the‌ cellular edge, at no‌ cost for mobile operators.‌‌ This work was accepted and presented at the‌ ACM Asia Conference on‌ Computer and Communications Security‌‌ (AsiaCCS 2025) 31. Beyond its scientific contribution,‌ the results demonstrate a‌ practical, standards-aware solution for‌‌ real-time SIMBox fraud prevention at the cellular edge,‌ with direct applicability for‌ mobile network operators. Based‌‌ on this work, an Inria Startup Studio maturation‌ procedure was initiated in‌ 2025 to explore technology‌‌ transfer and industrial valorization perspectives.

8.8 [Axis 3]‌ Decentralized network mechanisms and‌ architectures

8.8.1 ANSB: An‌‌ Optimized Network Slicing Scheme for Adaptive Load Balancing‌ in 5G Core Network‌

Participants: Lam Thanh-Son Nguyen‌‌, Nadjib Aitsaadi [UVSQ Paris-Saclay & DAVIDLab, France]‌, Cédric Adjih.‌

As 5G technology is‌‌ widely adopted, enterprises seek solutions for automation and‌ rapid service delivery. Network‌ Slicing (NS) leverages 3GPP‌‌ standards to create multiple, customized network slices on‌ shared infrastructure, serving diverse‌ applications and user groups.‌‌ This paper focuses on 3GPP 5G Core NS,‌ particularly Release 17, and‌ proposes Adaptive Network Slice‌‌ Balancing (ANSB) to optimize resource utilization by adjusting‌ User Equipment (UEs) and‌ Protocol Data Unit (PDU)‌‌ sessions. Extensive experimentation, with 5G OpenAirInterface (OAI) testbed,‌ demonstrates signiﬁcant improvements in‌ UEs, PDU sessions, and‌‌ maximize overall data rate consumption.

This article was‌ presented at the conference‌ ICC 2025 34.‌‌

8.8.2 +Tour: Recommending personalized itineraries for smart tourism‌

Participants: João Paulo Esper‌ [Federal University of Goiás,‌‌ Brazil], Luciano de S. Fraga [Federal University‌ of Goiás, Brazil],‌ Aline Carneiro Viana,‌‌ Kleber Vieira Cardoso [Federal University of Goiás, Brazil]‌, Sand Luz Correa‌ [Federal University of Goiás,‌‌ Brazil].

Next-generation touristic services will rely on‌ the advanced mobile networks'‌ high bandwidth and low‌‌ latency and the Multi-access Edge Computing (MEC) paradigm‌ to provide fully immersive‌ mobile experiences. As an‌‌ integral part of travel planning systems, recommendation algorithms‌ devise personalized tour itineraries‌ for individual users considering‌‌ the popularity of a city's Points of Interest‌ (POIs) as well as‌ the tourist preferences and‌‌ constraints. However, in the context of next-generation touristic‌ services, recommendation algorithms should‌ also consider the applications‌‌ (e.g., social network, mobile‌ video streaming, mobile augmented reality) the tourist will‌ consume in the POIs and the quality in‌ which the MEC infrastructure will deliver such applications.‌ In this paper, we address the joint problem‌ of recommending personalized tour itineraries for tourists and‌ efficiently allocating MEC resources for advanced touristic applications.‌ We formulate an optimization problem that maximizes the‌ itinerary of individual tourists while optimizing the resource‌ allocation at the network edge. We then propose‌ an exact algorithm that quickly solves the problem‌ optimally, considering instances of realistic size. Using a‌ real-world location-based photo-sharing database, we conduct and present‌ an exploratory analysis to understand preferences and users'‌ visiting patterns. Using this understanding, we propose a‌ methodology to identify user interest in applications.

Finally,‌ we evaluate our algorithm using this dataset. Results‌ show that our algorithm outperforms a modified version‌ of a state-of-the-art solution for personalized tour itinerary‌ recommendation, demonstrating gains up to 11% for resource‌ allocation efficiency and 40% for user experience. In‌ addition, our algorithm performs similarly to the modified‌ state-of-the-art solution regarding traditional itinerary recommendation metrics.

This‌ work was publised at the Computer Networks Elsevier‌ journal in 2025 13.

8.8.3 BUBBLE-BLUE a‌ multihop private network based on Bluetooth

Participants: Philippe‌ Jacquet, Nadjib Achir.

The aim of‌ the project is to create a kind of‌ “terrestrial STARLINK™” network based on users’ smartphones. The‌ BUBBLE-BLUE (BB) project aims to create private Bluetooth‌ bubbles on top of smartphones. In each private‌ bubble, participants will be able to communicate autonomously,‌ without recourse to private operator networks, neither data‌ nor cellular, relying solely on the Bluetooth technology‌ of smartphones. The rout- ing strategy is based‌ on dynamic Connected Dominant Sets (CDS). We present‌ the specific features of a BB network as‌ well as some simulation results on their routing‌ performance.

A preprint describing the protocol has been‌ prepared and is available in 46

8.9 [Axis‌ 3] Machine Learning enhanced network protocols and classical‌ network optimization (methods and techniques)

8.9.1 Precise Regularized‌ Minimax Regret with Unbounded Weights

Participants: Michael Drmota‌ [TU-Wien, Austria], Philippe Jacquet, Changlong Wu‌ [ University of Arizona, USA], Wojciech Szpankowski‌ [Purdue University, USA].

In online learning, a‌ learner receives data in rounds and, at each‌ round, predicts a label that is then compared‌ to the true label, incurring a loss. The‌ total loss over T rounds, when compared to‌ the loss of the best expert from a‌ class of experts or forecasters, is called the‌ regret. In this paper, we focus on logarithmic‌ loss for logistic-like experts with unbounded ddimensional weights,‌ a scenario that has been largely unexplored.

To‌ address the irregularities introduced by the unbounded weight‌ norm, we introduce a regularized version of the‌ average (fixed design) minimax regret by imposing a‌ soft constraint on the weight norm. We demonstrate‌ that the regularized minimax regret is fully characterized‌ by a complexity measure we term the regularized Shtarkov sum. We also‌ show how the behavior‌ of the standard regret‌‌ can be inferred from the regularized regret. Our‌ main results provide a‌ precise characterization of the‌‌ regularized Shtarkov sum and, consequently, the regularized regret‌ with unbounded weights up‌ to second-order asymptotics. Notably,‌‌ unlike the d/2 log T regret growth known‌ for bounded weights, our‌ results imply that the‌‌ regularized regret grows as $(\frac{1}{2} + \frac{α‌}{4}) d log T‌$ when the regularization parameter‌‌ is of order $Θ (T^{- α})$ .‌ We achieve this using‌ tools from analytic combinatorics,‌‌ including multidimensional Fourier analysis, the saddle point method,‌ and the Mellin transform.‌

8.9.2 Time- and Latency-Aware‌‌ Machine Learning for Vehicular Networks: Fast Federated Training‌ and Device-Adaptive Model Selection‌

Participants: Lucas Airam Castro‌‌ de Souza, Matteo Sammarco [Federal University of‌ Rio de Janeiro, Brazil]‌, Nadjib Achir,‌‌ Miguel Elias Mitre Campista [Federal University of Rio‌ de Janeiro, Brazil],‌ Luís Henrique Maciel Kosmalski‌‌ Costa [Federal University of Rio de Janeiro, Brazil]‌.

Vehicular networks are‌ increasingly adopting machine learning‌‌ to support intelligent, safety-critical applications. These applications face‌ strict challenges, including low‌ latency, high mobility, intermittent‌‌ connectivity, and diverse on-board devices. In this work,‌ we address two complementary‌ challenges: First, how to‌‌ reduce the end-to-end training delay in highly dynamic‌ vehicular networks? and second‌ (ii) how to select‌‌ models that can meet application latency requirements across‌ various automotive hardware characteristics‌ and capabilities.

We first‌‌ introduce time-optimized federated learning strategy (TOFEL) for vehicular‌ threat detection. TOFEL relies‌ on a client selection‌‌ mechanism that prioritizes participants expected to minimize overall‌ training time to fail.‌ Our results show that‌‌ carefully selecting only 20% of the available clients‌ can reduce the time‌ to reach high accuracy‌‌ by up to 50% compared to state-of-the-art solutions,‌ while also lowering resource‌ consumption on client devices.‌‌

Second, we propose a model selection system designed‌ specifically for automotive applications.‌ This system estimates inference‌‌ latency and determines whether potential models meet application‌ deadlines across various device‌ types. By utilizing both‌‌ device characteristics and model features, it predicts the‌ inference time for each‌ sample and filters out‌‌ models that do not meet latency requirements. This‌ approach enables reliable deployment‌ in diverse environments. To‌‌ ensure accurate estimation of inference time, the system‌ compares a deterministic analytical‌ model with four generative‌‌ AI-based estimators, using real execution measurements as the‌ ground truth.

This work‌ was accepted to be‌‌ published at Brazilian symposium SBRC 2025 38.‌

8.9.3 Topology optimization in‌ mobile wireless networks using‌‌ machine learning

Participants: Félix Marcoccia [Inria Paris, Paris;‌ Thales SIX, Gennevilliers, France;‌ Sorbonne Université, Paris, France]‌‌, Victor Fagoo [Thales SIX, Gennevilliers, France],‌ Gilles Monzat de Saint‌ Julien [Thales SIX, Gennevilliers,‌‌ France], Cédric Adjih, Thomas Watteyne [Inria‌ Paris, AIO], Paul‌ Mühlethaler [Inria Paris, AIO]‌‌.

This thesis 43 was defended on Oct.‌ 13, 2025. It extensively‌ studied modern machine learning‌‌ techniques (diffusion on graphs,‌ partial diffusion, VAE, DRL, etc.) and modern model‌ architectures (Transformers, GNN, Graph Transformers, GAT, GCN, etc.)‌ and improvements (U-shape, register tokens, CAM tokens, directional‌ density encodings, etc.), on the combinatorial problem of‌ selecting antenna directions for a multi-hop wireless ad‌ hoc network. In detail:

Mobile aerial networks have‌ emerged as compelling technologies due to their capacity‌ to deliver autonomous, infrastructure-free communication in dynamic environments.‌ Their growing relevance is driven by a wide‌ range of practical applications, ranging from UAVs to‌ planes and satellites. In order to overcome the‌ need for a centralized proxy, to achieve higher‌ resilience and capacity, such networks can leverage ad‌ hoc, multi-hop communications between nodes. However, they generally‌ suffer from theoretical limitations, particularly when using omnidirectional‌ antennas. To overcome these limitations and leverage directional‌ antennas, it becomes necessary to orchestrate all antenna‌ steering decisions, transmissions and receptions in real time,‌ ensuring a viable and efficient network topology. Given‌ the highly combinatorial nature of this problem, this‌ thesis proposes to address it using artificial intelligence‌ techniques, including supervised learning and generative models. In‌ the course of this thesis, we experiment with‌ various deep learning methods to solve our problem‌ and develop several solution architectures. By adapting and‌ extending state-of-the-art deep learning methods, we propose a‌ data-driven method which generates high-performance network configurations in‌ real time. Furthermore, leveraging advanced generative approaches, we‌ propose a learning architecture capable of jointly generating‌ the network links and a compatible transmission schedule,‌ while accounting for the network's dynamic behavior. The‌ resulting models yield a substantial theoretical throughput improvement‌ over conventional omnidirectional protocols, with even better scalability‌ as the number of nodes increases.

8.9.4 TopoFormer:‌ An Efficient Link-Set Prediction Architecture for Ad Hoc‌ Network Topology Generation

Participants: Félix Marcoccia [Inria Paris,‌ Paris; Thales SIX, Gennevilliers, France; Sorbonne Université, Paris,‌ France], Victor Fagoo [Thales SIX, Gennevilliers, France]‌, Gilles Monzat de Saint Julien [Thales SIX,‌ Gennevilliers, France], Cédric Adjih, Thomas Watteyne‌ [Inria Paris, AIO], Paul Mühlethaler [Inria Paris,‌ AIO].

In this paper, we present TopoFormer,‌ a powerful architecture for predicting links between communication‌ nodes in mobile networks. The goal is to‌ imitate, in real time, the results of a‌ costly combinatorial algorithm that generates topologies for networks‌ with directional antennas. These antennas offer excellent performance‌ but require complex, interdependent steering decisions in real‌ time. Our Transformer-based architecture is enhanced with efficient‌ components that add useful inductive biases, making it‌ suitable for environments where scaling is limited. A‌ key contribution is the introduction of directional density‌ encodings, which help the attention mechanism better separate‌ nodes in dense clusters. Equipped with our modules,‌ a single Transformer block of dimension 12 achieves‌ over 95 % accuracy, reducing the gap to‌ optimality by half compared to a plain 1-block‌ Transformer while requiring only 12 % more computation.‌ Using two blocks, the model comes close to‌ perfect accuracy.

This article 52 was presented at the Fourth Learning on‌ Graph Conference (LoG) 2025.‌

8.9.5 DSCAN-Net: A Dual-Stream‌‌ Network for Classifying Modulation Schemes in OTFS Systems‌

Participants: Tonmoy Rajkhowa [IIT‌ (BHU) Varanasi], Amit‌‌ Singh [IIT (BHU) Varanasi], Sanjeev Sharma [IIT‌ (BHU) Varanasi], Kuntal‌ Deka [IIT Guwahati],‌‌ Cédric Adjih.

Sixth-generation (6G) wireless communication systems‌ are required to ensure‌ seamless connectivity among devices‌‌ in high-mobility scenarios. In such scenarios, these systems‌ must prevail over the‌ challenges imposed by inter-carrier‌‌ and intersymbol interferences arising in doubly-spread channels (delay‌ and Doppler shifts) with‌ low-transmit power. Orthogonal timefrequency‌‌ space (OTFS) has emerged as a promising solution‌ in doubly-spread channels. Further,‌ classifying the modulation schemes‌‌ present in received signals accurately is very challenging‌ under low signal-to-noise ratio‌ (SNR) conditions. To address‌‌ these challenges, this work proposes DSCAN-Net: a dual-stream‌ crossattention network that leverages‌ the advantages of temporal‌‌ in-phase/quadrature (I/Q) and spatial amplitude/phase (A/P) formats of‌ the received OTFS signal‌ in two separate streams.‌‌ DSCAN-Net combines a convolutional neural network (CNN) and‌ residual channel attention network‌ (RCAN) to extract meaningful‌‌ representations from these two formats separately. These representations‌ are mapped with each‌ other using multiheaded cross-attention‌‌ to enhance the cross-context representation learning that improved‌ the overall classification accuracy‌ to 72.92%. Experimental results‌‌ demonstrate the effectiveness of the proposed DSCAN-Net under‌ low SNR conditions and‌ claim superior accuracy over‌‌ existing approaches utilized in OTFS systems. Additionally, the‌ results also signify the‌ effectiveness of A/P formats‌‌ over I/Q in modulation classification.

This article 35‌ was presented at NCC‌ 2025, in March 2025.‌‌

8.9.6 Linear programming for UAVs search path planning‌ in livestock health monitoring‌

Participants: Najoua Benalaya [ENSI,‌‌ Tunisia, University of Manouba], Ichrak Amdouni [ENSI,‌ Tunisia, University of Manouba]‌, Cédric Adjih [Inria‌‌ Saclay, France], Anis Laouiti [Telecom SudParis, France]‌, Leila Azouz Saidane‌ [ENSI, Tunisia, University of‌‌ Manouba].

UAV-Assisted Livestock Monitoring is a highly‌ relevant and essential application.‌ It involves deploying autonomous‌‌ Unmanned Aerial Vehicles (UAVs) to gather remote information‌ from various sensors and‌ IoT devices attached to‌‌ the livestock's necks. Such information includes the health‌ status indicators of the‌ cattle like temperature, respiration‌‌ rate, images or videos of the activity, etc.‌ The practical implementation of‌ this application presents several‌‌ challenges. One significant obstacle is the lack of‌ accurate cattle position information.‌ Employing the Global Positioning‌‌ System (GPS) has limitations like the high cost,‌ and the need for‌ a reliable network connection,‌‌ which may not be available in all rural‌ areas. Even using passive‌ tags like RFID tags‌‌ is not very practical due to their limited‌ reading distance. Thus, the‌ imperfect knowledge of the‌‌ cattle location forces the UAV to perform area‌ exploration and cattle searches.‌ The focus of this‌‌ research work is to design a model that‌ determines the optimal UAV‌ search path to localize‌‌ cattle.

We denote this issue as UAV Cattle‌ Search (UCS) path planning.‌ In a previous work,‌‌ we addressed the UCS‌ problem assuming a single stationary cattle (denoted UCS-ST‌ problem). We now extend this problem with two‌ new assumptions : (i) a single moving cattle‌ (UCS-SMT problem), and (ii) two moving cattle (UCS-TMT‌ problem). For each of these problems, we elaborate‌ a Mixed-Integer Linear Programming formulation (MILP) where the‌ objective function is the total expected search time.‌

Minimizing the search time is crucial for successful‌ search missions. However, to the best of our‌ knowledge, the literature did not focus on finding‌ the fastest path while guaranteeing the target localization.‌ Thus, in the conducted work, we focused on‌ the time required for a UAV to locate‌ a target and formulated an objective function aiming‌ at reducing this time. We implemented the models‌ using mathematical optimization software. Running different instances, our‌ models find optimal solutions that guarantee accurate cattle‌ localization while minimizing the expected search time for‌ graphs including up to 36 vertices (UCS-ST). We‌ have been inspired by established formulations in the‌ literature addressing related problems such as the Travelling‌ Salesman Problem and Optimal Search path. However, to‌ the best of our knowledge, the exact linear‌ formulations of our specific problems have never been‌ proposed.

This article 11 has been accepted in‌ the journal Computers and Electronics in Agriculture.

8.10‌ [Axis 3] Edge network offloading (methods and techniques)‌

8.10.1 Vehicular Cloud Computing as a Cost-Effective, Profitable,‌ and Sustainable Alternative to 5G Edge Computing

Participants:‌ Rosario Patanè [University Paris-Saclay, France], Andrea Araldo‌ [Télécom SudParis], Nadjib Achir, Lila Boukhatem‌ [University Paris-Saclay, France].

Edge Computing (EC) is‌ a computational paradigm that involves deploying resources such‌ as CPUs and GPUs near end-users, enabling low-latency‌ applications like augmented reality and real-time gaming. However,‌ deploying and maintaining a vast network of EC‌ nodes is costly, which can explain its limited‌ deployment today. A new paradigm called Vehicular Cloud‌ Computing (VCC) has emerged and inspired interest among‌ researchers and industry. VCC opportunistically utilizes existing and‌ idle vehicular computational resources for external task offloading.‌

In this work we are the first to‌ systematically address the following question: Can VCC replace‌ EC for low-latency applications? Answering this question is‌ highly relevant for Network Operators (NOs), as VCC‌ could eliminate costs associated with EC given that‌ it requires no infrastructural investment. Despite its potential,‌ no systematic study has yet explored the conditions‌ under which VCC can effectively support low-latency applications‌ without relying on EC. This work to fills‌ that gap.

Extensive simulations allow for assessing the‌ crucial scenario factors that determine when this EC-to-VCC‌ substitution is feasible. Considered factors are load, vehicles‌ mobility and density, and availability. Potential for substitution‌ is assessed based on multiple criteria, such as‌ latency, task completion success, and cost. Vehicle mobility‌ is simulated in SUMO, and communication in NS3‌ 5G-LENA. The findings show that VCC can effectively‌ replace EC for low-latency applications, except in extreme‌ cases when the EC is still required (latency < 16 ms).

Finally,‌ to ensure VCC is‌ not only feasible but‌‌ also adoptable, we also introduced, in a second‌ work, a comprehensive management‌ framework that jointly optimizes‌‌ latency, energy consumption, monetary incentives, and carbon emissions.‌ It relies on an‌ energy-aware task allocation strategy‌‌ that maximizes aggregate stakeholder utility while meeting deadlines‌ and minimizing energy costs,‌ coupled with a game-theoretic‌‌ revenue-sharing mechanism tailored to dynamic vehicular environments to‌ preserve incentives even when‌ some participants contribute weakly‌‌ or negatively. Simulations demonstrate that this approach sustains‌ low-latency execution, enables effective‌ monetization of vehicular resources,‌‌ and can reduce CO₂ emissions by more than‌ 99% compared to conventional‌ edge infrastructures—positioning VCC as‌‌ a cost-effective, profitable, and sustainable alternative to edge‌ computing in 5G networks.‌

Part of this work‌‌ was accepted to be published at Computer Networks‌ Journal in 2025 16‌. The second part‌‌ is an ongoing submission in ICDCS 2026. Rosario‌ Patanè defended his thesis‌ in Nov. 2025 44‌‌

8.10.2 Edge AI

Participants: Cédric Adjih, Nadjib‌ Achir, Amira Dhaouadi‌, Emmanuel Baccelli,‌‌ Yijie Luo, Fernando Molano, Mehdi Debbah‌, Yunmeng Shu,‌ Pengwenlong Gu.

This‌‌ year, work continued on the topic of EdgeAI;‌ in particular, we continued‌ to work on the‌‌ novel technique for embedded IoT systems that uses‌ support from edge or‌ cloud servers, and we‌‌ proposed a split-computing model. We also experimented with‌ developing models on Nvidia‌ Jetson Nano/Xavier/Orin embedded AI‌‌ boards, for specific applications, and continued develop an‌ testbed system.

8.10.3 Enhancing‌ Split ViT Inference Through‌‌ Sparsity-Driven Compression

Participants: Amida Dhaouadi, Nadjib Achir‌, Cédric Adjih.‌

Vision Transformer (ViT) is‌‌ a deep learning model that plays a significant‌ role in advanced computer‌ vision and pattern recognition‌‌ tasks but faces challenges in inference due to‌ high computational costs and‌ energy consumption. Split computing,‌‌ which distributes the load between an edge server‌ and a mobile device,‌ is proposed to address‌‌ these issues. Unfortunately, most literature focuses on split‌ computing for CNNs, with‌ limited attention to ViT‌‌ split computing. In this paper, we explore the‌ use of split computing‌ to optimize ViT's inference‌‌ process by limiting usage of bandwidth. We propose‌ compressing the latent space‌ data by introducing more‌‌ sparsity in the intermediate features. This sparsity is‌ then exploited through a‌ compression algorithm before transmitting‌‌ the data through a communication channel. To understand‌ and explain the performance‌ of our approach, we‌‌ analyze the latent space data using several metrics.‌ Our approach obtains a‌ significant compression ratio without‌‌ causing a substantial decrease in accuracy. It is‌ computationally efficient and do‌ not require retraining the‌‌ model.

This article was presented at VTC 2025‌ 24.

8.11 [Axis‌ 3] Security of the‌‌ edge/core compound including IoT deployments (technologies)

8.11.1 Delay‌ analysis of BFT consensus‌ : Case study of‌‌ Narwhal and Bullshark protocols

Participants: Khouloud Hwerbi [ENSI,‌ Tunisia, University of Manouba]‌, Ichrak Amdouni [ENSI,‌‌ Tunisia, University of Manouba]‌, Cédric Adjih, Leila Azouz Saidane [ENSI,‌ Tunisia, University of Manouba], Anis Laouiti [Telecom‌ SudParis, France].

Acknowledging the critical influence of‌ consensus delays on blockchain performance, this paper presents‌ an analytical and simulation-based exploration of delay characteristics‌ in Byzantine Fault Tolerant (BFT) consensus mechanisms. Our‌ focus is on SUI, a blockchain system that‌ employs a Directed Acyclic Graph (DAG) structure to‌ support parallel transaction execution. SUI relies on two‌ integrated protocols: Narwhal, a mempool protocol responsible for‌ efficient block dissemination and DAG construction; and Bullshark,‌ which organizes DAG vertices to produce a consistent‌ total order of transactions without incurring additional communication‌ overhead.

While our previous work modeled Narwhal's delay‌ characteristics under various message propagation distributions, this study‌ shifts attention to Bullshark—the protocol responsible for reaching‌ consensus. We propose a probabilistic analytical model that‌ estimates the number of rounds required to reach‌ consensus. In this model, each validator's decision is‌ treated as a Bernoulli trial, and we apply‌ the binomial distribution to determine the probability of‌ reaching quorum. This framework enables us to analyze‌ the expected delay of the protocol.

To validate‌ our model, we implemented both Narwhal and Bullshark‌ and conducted extensive simulations. The simulation results show‌ strong agreement with our analytical predictions, confirming the‌ accuracy of our model. For instance, under a‌ Gaussian delay model with mean and standard deviation‌ ms—values representative of short-range wireless communication in real-world‌ IoT or LAN settings [1]—we predict an average‌ round duration of approximately 3.26 ms. Furthermore, based‌ on our binomial-based model of block commitment, the‌ expected number of rounds to reach consensus is‌ approximately 1 when , indicating that blocks typically‌ commit in a single round with high probability.‌

To the best of our knowledge, this is‌ the first study to model Bullshark's consensus process‌ using Bernoulli trials and binomial distributions. Our contributions‌ offer a novel framework for evaluating its efficiency‌ and provide insights that can guide future optimization‌ and scalability efforts for DAG-based BFT protocols.

This‌ article 14 is published in Computer Communications, October‌ 2025.

8.11.2 Delays in mempool-based blockchains under realistic‌ conditions: case of Narwhal and DroneVet application

Participants:‌ Khouloud Hwerbi [ENSI, Tunisia, University of Manouba],‌ Ichrak Amdouni [ENSI, Tunisia, University of Manouba],‌ Cédric Adjih, Leila Azouz Saidane [ENSI, Tunisia,‌ University of Manouba], Anis Laouiti [Telecom SudParis,‌ France].

This study focuses on modeling and‌ analyzing the delay in consensus protocols within mempool-based‌ blockchains. In such systems, unconfirmed transactions are temporarily‌ stored in a memory pool (mempool) before being‌ selected, ordered, and committed to blocks by the‌ consensus mechanism. In a previous work, we designed‌ a generic model of a mempool-based blockchain, capturing‌ core protocol behaviors. The model involves n nodes‌ (potentially including Byzantine nodes) that exchange blocks and‌ await one or two quorums of acknowledgments before‌ producing new blocks and advancing to the next‌ round. Based on this framework, we derived two Markov chains to characterize‌ round durations and validated‌ them through simulations.

In‌‌ this paper, we extend that work in three‌ key directions: (1) We‌ focus on Narwhal, a‌‌ specific mempool protocol, describing its operations and its‌ round duration formula. (2)‌ We introduce DroneVet, an‌‌ application leveraging UAVs to monitor livestock health and‌ shares collected data across‌ stakeholders via blockchain technology.‌‌ DroneVet is implemented using Narwhal and includes realistic‌ environmental constraints such as‌ packet loss, network latency,‌‌ and UAV-energy limitation. (3) We develop DroneVet and‌ compare its round duration‌ with the theoretical round‌‌ duration of Narwhal. By taking into account the‌ environmental factors specific to‌ agricultural deployments, this work‌‌ highlights how such conditions impact the reliability and‌ responsiveness of blockchain-based systems.‌ Our findings provide valuable‌‌ insights for deploying secure, resilient, and transparent blockchain‌ solutions in rural settings‌ to support reliable livestock‌‌ monitoring.

This article 15 is accepted for publication‌ in Annals of Telecommunications‌ 2026.

8.11.3 Military IoT‌‌ from Management to Perception: Challenges and Opportunities Across‌ Layers

Participants: Paulo Rettore‌ [Fraunhofer FKIE, Germany],‌‌ Jannis Mast [Osnabrück University , Germany], Thorsten‌ Aurisch [Fraunhofer FKIE, Germany]‌, Aline Carneiro Viana‌‌, Peter Sevenich [Fraunhofer FKIE, Germany], Bruno‌ Santos [Federal University of‌ Bahia, Brazil].

The‌‌ dynamic and challenging nature of battlefield scenarios necessitates‌ robust, secure, and e!cient‌ networking systems to connect‌‌ and share context-aware information timely, which is crucial‌ for achieving precision and‌ mission success. The Military‌‌ Internet of Things (MIoT) has the potential to‌ fulfill these demands holistically‌ and coherently. MIoT is‌‌ pivotal in creating a networked environment where interconnected‌ heterogeneous devices enhance situational‌ awareness, communication, and decisionmaking.‌‌ Yet, dynamic conditions (e.g., device heterogeneity, mobility, unreliable‌ wireless communication, operation scenarios,‌ etc.) impose challenges across‌‌ di"erent layers in common coalition domains.

Our article‌ 17 publised at the‌ IEEE Internet of Things‌‌ Magazine 2025 discusses these challenges, current literature, and‌ future design advances related‌ to security/privacy, mobility, robustness,‌‌ and data-driven AI solutions in potential MIoT applications.‌

9 Bilateral contracts and‌ grants with industry

9.1‌‌ Bilateral grants with industry

Thalès - CIFRE Thesis‌

Participants: Cedric Adjih,‌ Paul Muhlethaler, Felix‌‌ Marcoccia.

Felix Marcoccia is a CIFRE student‌ at Thalès, co-advised at‌ Inria by P. Mühlethaler‌‌ and C.Adjih, on the subject of: "Study of‌ MANET Solutions for a‌ Radio Communication System Based‌‌ on Artificial Intelligence Algorithms". He defended in October‌ 2025.

Qualcomm - Donation‌

Participants: Emmanuel Baccelli,‌‌ Philippe Jacquet.

We have finalized a donation‌ process from Qualcomm industry,‌ starting year 2024 and‌‌ supporting the research on wireless IoT and routing,‌ in particular the experimentation‌ of local wireless bubble‌‌ based on Bluetooth.

Fujitsu / RunMyProcess - Donation‌

Participants: Emmanuel Baccelli.‌

100,000 euros. We have‌‌ received this donation to support us in developing‌ and maintaining RIOT.

PADAM‌ Mobility - CIFRE Thesis‌‌

Participants: Aline Carneiro Viana, Mohamed Ourahou.‌

Mohamed Ourahou is a‌ CIFRE PhD student at‌‌ PADAM Mobility (Siemens Mobility‌ Group - Paris), co-advised at Inria by A.‌ Carneiro Viana, at Telecom SudParis by A. Araldo,‌ and at PADAM Mobility by L. Zigrand, on‌ the subject of: "Geostatistical and Machine Learning Methods‌ for Sustainable Deployment of Mobility on Demand".

SAFRAN-‌ CIFRE Thesis

Participants: Cedric Adjih, Paul Muhlethaler‌, Corentin Gautier.

Corentin Gautier is a‌ CIFRE PhD student at SAFRAN Electronics & Defense,‌ co-advised at Inria by P. Mühlethaler and C.‌ Adjih, on the subject of: "FANET for Vehicles‌ Swarms".

10 Partnerships and cooperations

10.1 International initiatives‌

10.1.1 Inria Berlin

Participants: Emamnuel Baccelli.

Emmanuel‌ Baccelli is the scientific director of Inria Berlin.‌ Inria Berlin is an initiative from Inria to‌ foster and increase its scientific collaborations with the‌ Berlin research and innovation ecosystem. As such, Inria‌ has entered into a partnership with the Einstein‌ Center for Digital Future (ECDF). Historically, several Inria‌ researchers have been well established in Berlin, engaging‌ in long-term interactions with prominent academic institutions from‌ the city across various research domains, including but‌ not limited to Internet of Things (IoT), Cybersecurity,‌ Digital Humanities and Open Science. In this context,‌ Inria and the ECDF signed a Memorandum of‌ Understanding (MoU) in September 2023 with the goal‌ of fostering scientific exchanges and developing joint research‌ projects in several fields related to digital sciences‌ in the Berlin-Potsdam area. Inria Berlin has since‌ fostered the emergence of several new Inria associate‌ teams. Up-to-date information is available on the Inria‌ Berlin website.

10.1.2 Participation in other International‌ Programs

PHC ANGEL 2024-

Participants: Cedric Adjih.‌

Title:
“Agriculture Numérique et `diGital twin' face aux‌ changements climatiques pour une sÉcurité aLimentaire” (PHC-Maghreb 2024)‌ [link]
Coordinator:
Telecom SudParis (France), ENSI/U.‌ of Manouba (Tunisia), ENSIAS (Morocco)
Partners:
Laboratoire CRISTAL,‌ ENSI Tunisia, Telecom SudParis, IPP, France Inria Saclay,‌ France. From TRiBE: Cedirc Adjih .
Description:‌
The project aims to enhance agricultural resilience and‌ sustainability against climate change and food security challenges‌ through advanced technologies like Digital Twin (DT), AI,‌ IoT, UAVs, and Blockchain.rance)

10.2 International research visitors‌

10.2.1 Visits of international scientists

Bernard Man

Status‌
Full Professor
Institution of origin:
Macquarie University
Country:‌
Australia
Dates:
From April 2025 until July 2025‌
Context of the visit:
On going collaboration on‌ efficient and energy saving blockchains and on hyperfractal‌ models 59
Mobility program/type of mobility:
Research stay‌ funded by Inria Saclay funding.

10.3 European initiatives‌

10.3.1 Other european programs/initiatives

TinyPART (2021–2024):

Participants: Emmanuel‌ Baccelli.

Title:
Tiny, PrivAte, pRoven and isolaTed‌ (ANR/BMBF French German Cybersecurity Program) [link]‌
Coordinator:
Orange.
Partners:
FU Berlin, Lille University, and‌ PHYSEC GmbH. From TRiBE: Emmanuel Baccelli
Description:‌
TinyPART develops Software-Defined IoT building blocks for low-power‌ devices, emphasizing privacy-by-design and cybersecurity. It enables isolating‌ untrusted IoT logic, integrating privacy-oriented preprocessing like differential‌ privacy and lightweight cryptography. Built on RIOT OS‌ and PIP, TinyPART explores tradeoffs between isolation, security,‌ memory footprint, and developer usability.
Team contribution:
novel designs for tiny software‌ containers and experimental platforms‌ for TinyML. Open-source implementations‌‌ of these designs were upstreamed to RIOT.

10.4‌ National initiatives

10.4.1 AAPs‌

IoT-LAB (now part of‌‌ SLICES-FR):

Participants: Cedric Adjih, Fernando Molano,‌ Emmanuel Baccelli.

Partners:‌
Sorbonne Université, Inria (Lille,‌‌ Sophia-Antipolis, Grenoble), INSA, Télécom Paris, Télécom SudParis, LSIIT‌ Strasbourg.
Abstract:

FIT (Future‌ Internet of Things) had‌‌ developed an experimental facility, a federated and competitive‌ infrastructure with international visibility‌ and a broad panel‌‌ of customers. It provides this facility with a‌ set of complementary components‌ that enable experimentation on‌‌ innovative services for academic and industrial users. The‌ project gave french internet‌ stakeholders a means to‌‌ experiment on mobile wireless communications at the network‌ and application layers thereby‌ accelerating the design of‌‌ advanced networking technologies for the future internet.

SLICES-FR‌ is a larger-scale ongoing‌ effort to provide such‌‌ platforms, a follow-up and much more.

One component‌ of the existing platforms‌ is the sets of‌‌ IoT-LAB testbeds (see the IoT-LAB web site).‌ These were motivated by‌ the observation that the‌‌ world is moving towards an “Internet of Things”,‌ in which most communication‌ over networks will be‌‌ between objects rather than people.

Project 5G-mMTC:

Participants:‌ Cedric Adjih, Alexandre‌ Abadie [Inria, SED],‌‌ Nadjib Achir, Fernando Molano, Emmanuel Baccelli‌.

Funding instrument:
AAP‌ - Plan de relance‌‌ « Souveraineté dans les réseaux de télécommunications afin‌ d'accélérer les applications de‌ la 5G » (France‌‌ Relance)
Project acronym:
5G-mMTC
Duration:
2021–2024
Partners:
Amarisoft,‌ EDF R&D, Fédération francaise‌ de cyclisme, Inria Saclay,‌‌ Institut Mines Telecom, IS2T, Sequans communications, Sparkling Tech,‌ Université de Versailles (UVSQ‌ Paris-Saclay), Webdyn
Website:
5g-mMTC‌‌ web site
Abstract:
The 5G-mMTC project aims to‌ provide software and hardware‌ tools for the rapid‌‌ implementation of a 5G solutions for the IoT.‌ Two use cases will‌ be implemented directly within‌‌ the framework of this project: one developed in‌ conjunction with the French‌ Cycling Federation (FFC), which‌‌ will enable real-time analysis of athletes' data and‌ their performances; the other‌ will be worked on‌‌ jointly with EDF, to enable real-time management of‌ the entire fleet of‌ existing heterogeneous sensors

Inria‌‌ Challenge on Federated Learning FedMalin:

Participants: Cedric Adjih‌, Nadjib Achir,‌ Aline Carneiro Viana.‌‌

Partners:
Inria Teams (ARGO, COATI, COMET, EPIONE, MAGNET,‌ MARACAS, NEO, SPIRALS, TRIBE,‌ WIDE).
Abstract:
FedMalin is‌‌ a research project that spans 10 Inria research‌ teams and aims to‌ push FL research and‌‌ concrete use-cases through a multidisciplinary consortium involving expertise‌ in ML, distributed systems,‌ privacy and security, networks,‌‌ and medicine. We propose to address a number‌ of challenges that arise‌ when FL is deployed‌‌ over the Internet, including privacy & fairness, energy‌ consumption, personalization, and location/time‌ dependencies. FedMalin will also‌‌ contribute to the development of open-source tools for‌ FL experimentation and real-world‌ deployments, and use them‌‌ for concrete applications in medicine and crowdsensing. The‌ FedMalin Inria Challenge is‌ supported by Groupe La‌‌ Poste, sponsor of the‌ Inria Foundation.

10.4.2 ANR

QUANTINT

Participants: Philippe Jacquet‌.

Funding instrument/scientific committee:
PRCI
Project acronym:
QUANTINT‌
Project title:
Quantum Information and Network Theory: Algorithms‌ and Performance Limits
Duration:
2025–2028
Coordinator:
Philippe Jacquet‌
Other partners:
PHIQUS/Inria, EURECOM, SUNY Albany (US), University‌ of Michigan (US).
Budget:
1,112M€, TRiBE (143K€)
Abstract:‌
QUANTINT envisions an interconnected network of quantum devices‌ exchanging qubits and utilizing unique quantum properties such‌ as entanglement to enhance information processing algorithms. Towards‌ this vision, QUANTINT will design efficient universal algorithms‌ and strate- gies to (i) compress distributed qubits‌ and (ii) harness distributed entanglement in emerging tasks‌ such as distributed learning

MITIK

Participants: Aline Carneiro‌ Viana, Nadjib Achir, Abhishek Mishra,‌ Catuscia Palamidessi, Fernando Molano.

Funding instrument/scientific‌ committee:
PRC/CE25
Project acronym:
MITIK
Project title:
Mobility‌ and contact traces from non-intrusive passive measurements
Duration:‌
2020–2025
Coordinator:
Aline Carneiro Viana
Other partners:
COMETE/Inria,‌ Universite de la Rochelle, Sorbonne Universite (UPMC).
Budget:‌
644K€, TRiBE (289K€)
Web link:
ANR MITIK website‌
Abstract:
The MITIK project is a 42-month ANR‌ project that will start in February 2020. Mitik's‌ primary objective is the design of an entirely‌ new methodology to help the community obtain real‌ wireless contact traces that are non-intrusive, representative, and‌ independent of third parties. The secondary outcome of‌ the project is be the public release of‌ (1) the measurement tool designed for the easy‌ contact gathering task; (2) contact traces which are‌ clean, processed, and privacy-preserving, i.e., protecting both the‌ anonymity and the location privacy of the users;‌ and (3) their spatiotemporal statistical analysis. We expect‌ that Miti's outcomes will support non-biased research on‌ the modeling as well as on the leveraging‌ of wireless contact patterns.

PEPR NF FITNESS

Participants:‌ Aline Carneiro Viana, Cedric Adjih, Nadjib‌ Achir, Emmanuel Baccelli, Amira Dhaouadi.‌

Funding instrument/scientific committee:
PEPR Networks of the Future‌ - ANR
Project acronym:
NF FITNESS
Project title:‌
From IoT breakthroughs to Network Enhanced ServiceS
Duration:‌
2023–2030
Coordinator:
Eric Mercier (CEA)
Inria co-pilot:
Nadjib‌ Achir (TRiBE, Inria)
Other partners:
IMT, CNRS, Inria‌ (AGORA, AIO, FUN, TRiBE)
Budget:
4.9M€, Inria (900K€),‌ TRiBE (290K€)
Web link:
NF FITNESS website
Abstract:‌
The FITNESS project aims to provide elementary blocks‌ and define the conditions for their integration into‌ vertical applications with a guarantee of coexistence for‌ IoT. Three areas are addressed: Massive IoT (low‌ consumption and low cost), Industry 4.0 (Mission Critical‌ connectivity), and Vehicular and Connected Transport (towards Autonomous‌ Mobility). The key elements to consider are the‌ evolution towards standard protocols and the general coexistence‌ of new networks post-5G. Indeed, factories and manufacturing‌ centers are attentive and eager to evolve toward‌ digitization and wireless connectivity. However, robustness and the‌ ability to perform critical missions will be crucial.‌ In parallel, new services include digital twins and‌ connected and autonomous mobility. Therefore, it is essential‌ to ensure connectivity and access to safe, permanent,‌ and guaranteed resources. The NF-FITNESS will address the challenges raised by these‌ three main domains. The‌ research will include PHY,‌‌ NETWORK, and APPLICATION layers to generate outcomes tailored‌ to specific verticals. The‌ collaboration aims to:
- Enhance‌‌ the performance of foundational components, serving as a‌ foundational application for Massive‌ IoT, focusing on seamless‌‌ integration.
- Investigate the unique requirements of Mission Critical‌ applications, prioritizing robustness as‌ the most critical factor.‌‌
- Foster the development of resource sharing and interoperability,‌ emphasizing the challenges associated‌ with data processing.

PEPR‌‌ NF NAI

Participants: Aline Carneiro Viana, Nadjib‌ Achir.

Funding instrument/scientific‌ committee:
PEPR Networks of‌‌ the Future - ANR
Project acronym:
NF NAI‌
Project title:
Architectures and‌ Infrastructures de Réseaux et‌‌ Convergence réseaux, cloud and capteurs
Duration:
2023–2030
Coordinator:‌
Gérard Memmi (IMT)
Other‌ partners:
IMT, CNRS, EURECOM,‌‌ INP Toulouse, CentraleSupélec, INRIA (AGORA, DIANA, RESIST, TRIBE)‌
Budget:
5M€, Inria (490K€),‌ TRiBE (200K€)
Web link:‌‌
NF NAI website
Abstract:
Beyond traditional objectives (throughput,‌ execution speed, latency, object‌ connection density, etc.), the‌‌ NF-NAI project must allow the effective integration of‌ a multitude of new‌ technologies, such as those‌‌ of the physical layer (reconfigurable intelligent surfaces) or‌ the transition to 3D‌ (NTN – Non-Terrestrial Networks)‌‌ and architectural principles (such as slicing and end-to-end‌ dynamic orchestration). It must‌ facilitate the emergence of‌‌ new applications and services, thanks to transparency in‌ terms of performance, robustness,‌ and security with respect‌‌ to the use cases. The project will also‌ have to propose and‌ create interfaces with converged‌‌ network-cloud-sensing systems to offer a high degree of‌ transparency to developers of‌ applications ranging from the‌‌ edge to the cloud, from mini-connected objects to‌ large data centers through‌ Multi-access edge computing (MEC).‌‌

PEPR NF PERSEUS

Participants: Cedric Adjih, Paul‌ Mühlethaler.

Funding instrument/scientific‌ committee:
PEPR Networks of‌‌ the Future - ANR
Project acronym:
NF PERSEUS‌
Project title:
Power-Efficient Radio‌ interface for Sub-7GHz distributEd‌‌ massive MIMO infrastructUreS
Duration:
2023–2030
Coordinator:
Rafik Zayani‌ (CEA-Leti)
Other partners:
IMT,‌ CNRS, Inria (MARACAS, TRiBE,‌‌ EVA)
Budget:
5M€, Inria (300K€), TRiBE (70K€)
Web‌ link:
NF PERSEUS website‌
Abstract:
PERSEUS focuses on‌‌ the technologies, processing and optimization of cell-free massive‌ MIMO (CF-mMIMO) networks in‌ the sub-7 GHz frequency‌‌ band. CF-mMIMO technology, combined with reconfigurable intelligent surface‌ (RIS) techniques and artificial‌ intelligence (AI) tools, is‌‌ a highly promising solution for beyond-5G networks. PERSEUS‌ aims to increase the‌ maturity of these technologies‌‌ in order to achieve power- and spectrum-efficient massive‌ access. The project covers‌ several aspects with a‌‌ view to designing a "cell-free massive MIMO" network:‌ (i) design, manufacture and‌ test of RF circuits,‌‌ RIS and antennas, (ii) proposal of robust PHY‌ and MAC layers based‌ on signal propagation measurements‌‌ and the incorporation of hardware imperfection models, and‌ (iii) development of proofs‌ of concept to practically‌‌ evaluate the performance of the selected algorithms and‌ the hardware manufactured within‌ the framework of the‌‌ project.

PEPR NF FPNG

Participants: Cedric Adjih,‌ Fernando Molano.

Funding‌ instrument/scientific committee:
PEPR Networks‌‌ of the Future -‌ ANR
Project acronym:
NF FPNG
Project title:
French‌ Network of Test Platforms for the Next Generation‌ of Mobile Communications
Duration:
2023–2030
Coordinator:
Philippe Besnier‌ (CNRS)
Other partners:
IMT, EURECOM, CNRS, Sorbonne Université,‌ Inria (MARACAS, TRiBE, EVA)
Budget:
4.5M€, Inria (1.4M€),‌ TRiBE (157K€)
Web link:
NF FPNG website
Abstract:‌
The objective of the FPNG project is to‌ build a research infrastructure on a national scale‌ to test new hardware components and evaluate the‌ new paradigms of the next generation of telecommunications‌ networks. These research infrastructures target both core technology‌ components and end-to-end network testing. This platform program‌ aims to address all relevant technologies, ranging from‌ elementary electronic components to large-scale networking experiments, to‌ address all the specific challenges of the PEPR‌ Networks of the Future. The objective is to‌ grant the researchers of this PEPR free access‌ to existing infrastructures and to invest in new‌ strategic and advanced infrastructures when they still need‌ to be created to respond to the new‌ challenges.

PEPR MOBIDEC Mob Sci-Dat Factory

Participants: Aline‌ Carneiro Viana, Nadjib Achir, Philippe Jacquet‌, Lucas de Souza Felix, Geoffrey Deperle‌, Anne Josiane Kouam.

Funding instrument/scientific committee:‌
PEPR MOBIDEC (Data technology for Mobility in the‌ teriitories) - ANR
Project acronym:
Mob Sci-Dat Factory‌
Project title:
Sharing of tools for processing and‌ analysing mobility data
Duration:
2023–2027
Coordinator:
Aline Carneiro‌ Viana
Other partners:
UGE, IFPEN, IGN, CEREMA, Inria‌ (AGORA, ASCII, COATI, FUN, TRIBE)
Budget:
4 333‌ 114€ Inria (1 385 520,58€), TRiBE (766 500,24€)‌
Web link:
Mob Sci-Dat Factory website
Abstract:
Mob‌ Sci-Data Factory shares the PEPR's primary goal of‌ contributing to developing more sustainable mobility strategies by‌ providing decision-making support methodology and a digital toolbox‌ fed by appropriately selected and processed mobility data‌ and by a deeper understanding of the involved‌ transport uses and behaviors in mobility. This project‌ will clarify and extract the elements determining and‌ explaining the characteristics of mobility data, which also‌ raise the following questions:
- What data and what‌ are their availability, accessibility, quality, and representativeness?
- Which‌ methods and digital tools are necessary for processing,‌ calibrating, understanding, and enriching data while dealing with‌ missing data and new acquiring?
- What are the‌ specifications of the decision-support platform required for standard‌ tools and data research sharing?
Answering those three‌ questions together is a challenging task and the‌ primary goal of Mob Sci-Data Factory project. Mob‌ Sci-Dat Factory will make available in a secure‌ and privacy-compliant cloud-based infrastructure different sources of mobility‌ data together with open-source libraries and methods designed‌ to be unified, modular, and interoperable from conception.‌ Mob Sci-Dat Factory outcomes will facilitate data sovereignty‌ and open-source development interoperability across multiple scientific actors‌ in France, while accelerating research focused on mobility‌ by offering privacy-compliant and secure data accessibility

10.5‌ Regional initiatives

AI4Demand-Responsive Transit (2024–2027)

Participants: Aline Carneiro‌ Viana, Mohamed Ourahou, Andrea Araldo.‌

Title: Geostatistical and Machine Learning Methods for Sustainable Deployment of Mobility‌ on Demand (DIM AI4IDF‌ - Intelligence Artificielle centrée‌‌ sur l'humain en Ile de France)
Coordinator:‌ TPT-IPP
Partners: TRiBE,‌ PADAM Mobility. From TRiBE‌‌: A. Araldo (TST-IPP), A. Carneiro Viana, M.‌ Ourahou.
Grant: 3-y‌ PhD scholarship (Call: DIM‌‌ AI4IdF).
Description: Mobility on Demand (MoD) services‌ adapt vehicle routes to‌ user requests, focusing on‌‌ improving accessibility rather than just efficiency metrics. Accessibility‌ measures opportunities (jobs, schools,‌ shops) reachable within a‌‌ set time, promoting social, economic, and environmental sustainability.‌ It aims to leverage‌ Mobility on Demand to‌‌ reduce accessibility gaps between city centers and suburbs.‌
Team contribution: investigations‌ on how Demand-Responsive Transit‌‌ can improve accessibility in specific areas by evaluating‌ the brought additional opportunities‌ reachable within a limited‌‌ time frame.

11 Dissemination

11.1 Promoting scientific activities‌

11.1.1 Scientific events: organisation‌

General chair, scientific chair‌‌

Cédric Adjih :
- General co-chair of the 14th‌ IFIP/IEEE International Conference on‌ Performance Evaluation and Modeling‌‌ in Wired and Wireless Networks, PEMWN 2025,‌ which took place at‌ CNAM, Paris, November 25-27,‌‌ 2025.

Member of the organizing committees

Amira Dhaouadi‌ : Member of the‌ Junior Organization Committee of‌‌ the 10th Junior Conference on Data Sciences and‌ Engineering JDSE 2025 which‌ took place on 25-26‌‌ September 2025.
Aline Carneiro Viana : TPC co-chair‌ of IEEE/IFIP TMA 2025;‌
Aline Carneiro Viana ,‌‌ Nadjib Achir , and Anne Josiane Kouam :‌ Co-organizers of NetMob 2025‌, held in October‌‌ 2025 at CNAM, Paris, France. NetMob is a‌ leading international conference on‌ mobile data analysis and‌‌ modeling. Within the organization and together with other‌ members of the team‌ (C. Achir, F. Molano,‌‌ W. Viana, A. Bouroudi, G. Deperle), they served‌ in multiple key roles,‌ including:
- Data Challenge Co-organizers,‌‌
- TPC Co-Chair,
- Web Chair,
- Student Grant Chair,
- Registration‌ Chair,
- Publication Chair.
These‌ roles covered the scientific‌‌ coordination of the conference, the design and supervision‌ of the data challenge,‌ and the management of‌‌ the reviewing, publication, and dissemination processes.
Fernando Molano‌ , Wellington Viana Lobato‌ Junior , Amira Dhaouadi‌‌ : Essential part of the organization team of‌ PEMWN 2025, November 25-27,‌ 2025.
Cédric Adjih :‌‌ Member of the Senior Organization Committee of the‌ 10th Junior Conference on‌ Data Sciences and Engineering‌‌ JWOC 2025 which took place at Télécom Paris‌ on 3 October 2025.‌ Also Student Grant co-Chair‌‌ of NetMob 2025.

11.1.2 Scientific events: selection

Member‌ of the conference program‌ committees

Aline Carneiro Viana‌‌ : TPC member of NetMob 2025, Algotel 2025,‌ and IEEE TMA 2026;‌
Nadjib Achir : TPC‌‌ member of IEEE VCC 2025, PIMRC 2025, NetMob‌ 2025.
Cédric Adjih :‌ TPC member of ICC‌‌ 2025, ICC 2026, PEMWN 2025, and Reviewer of‌ ISIT 2025.
Anne Josiane‌ Kouam : TPC member‌‌ of WiMob 2025, PAM 2025, TMA 2025, Algotel&Cores‌ 2025 (national conference), and‌ AISec 2025.

11.1.3 Journal‌‌

Member of the editorial boards

Aline Carneiro Viana‌ :
- (Since 2024) Associate‌ editor of EPJ Data‌‌ Science;
- (Since 2014) Area‌ editor of ACM SIGCOMM Computer Communication Review (CCR).‌

Reviewer - reviewing activities

Nadjib Achir : Reviewer‌ for Annals of Telecommunications, Pervasive and Mobile Computing‌ Journal, Transaction on Mobile Computing.
Aline Carneiro Viana‌ : Reviewer of ACM SIGCOMM CCR, EPJ Data‌ Science.

11.1.4 Invited talks

Nadjib Achir : seminar‌ on LINCS Lab "Privacy-aware passive sniffing: from wireless‌ measurements to bounded trajectories", November 2025.
Cédric Adjih‌ : Talk at CNAM, Paris, 23 September 2025,‌ on "Modern Random Access for Grant-Free Cellular Networks".‌
Cédric Adjih : Keynote at the 3rd IoT&ET‌ Workshop (Sousse, Tunisia), October 21-23, 2025, on "‌On the Evolution of Modern Random Access for‌ Grant-Free Cellular Networks".
Cédric Adjih : presentation‌ of the NGC-AIoT platform at the first SLICES-FR‌ Summer School in Lyon (July 7-11, 2025).
Emmanuel‌ Baccelli : Talk on "Ariel OS - An‌ Open Source Embedded Rust OS for Networked Multi-Core‌ Microcontrollers" (RustWeek, May 2025).
Emmanuel Baccelli : Talk‌ on " Science et Numérique: Parcours de Recherche‌ en Cybersecurité" (Institut Français, Berlin, October 2025).
Aline‌ Carneiro Viana : talk on ”The Poetry of‌ Digital Presence: Human Beauty and Vulnerability behind Networking‌ Habits ” and on "A quick tour of‌ Inria" at Unicamp (State University of Campinas), Brazil‌ (July 30, 2025).
Aline Carneiro Viana : talk‌ on ”Understanding individuals' proclivity for novelty seeking” at‌ the UTFPR (Technological Federal University of Parana), Brazil‌ (July 24, 2025). This visit led to the‌ three-month internship of Gustavo Bruno dos Santos .‌
Aline Carneiro Viana : talk on ”A Privacy-Compliant‌ Passive Collection of WiFi Probe Request Datasets” at‌ the UFRJ (Federal University of Rio de Janeiro),‌ Brazil (August 12, 2025).

11.1.5 Scientific expertise

Cédric‌ Adjih served as an evaluation expert for the‌ ANR.
Nadjib Achir served as a project expert‌ for the CEFIPRA Program.

11.1.6 Research administration

Emmanuel‌ Baccelli Scientific Director of Inria Berlin.
Aline Carneiro‌ Viana is the leader of the TRiBE Project-Team‌ of Inria since its creation (July 2019)
Aline‌ Carneiro Viana was the coordinator of ANR MITIk‌ (since 2020-2025) and PEPR MOBIDEC Mob Sci-Data Factory‌ (PC3, 2023-2027) projects.
Nadjib Achir is the co-pilot‌ of PEPR NF FITNESS.
Cédric Adjih is part‌ of the Scientific Commission of Inria Saclay.
Cédric‌ Adjih is the co-animator of the COURSE (COmité‌ UtilisateuRs SlicEs-fr).

11.2 Teaching - Supervision -‌ Juries

11.2.1 Teaching

Cédric Adjih : 12h "Internet‌ of Things" lab sessions in 2025 at ENSEA.‌ Also supervision of Master Student projects in CNAM,‌ in 2025-2026.
Amira Dhaouadi : Teaching assistant in‌ Machine Learning for the CS department at LIX‌ (Ecole Polytechnique-IPP) (2024-2025).

11.2.2 Supervision

PhDs supervision (defended‌ in 2025):
- Rosario Patane , “VehiCloud: How can‌ Vehicles increase Cloud intelligence?”, started Dec. 2021, defended‌ on Nov. 2025. Advisors: Lila Boukhatem (Paris-Saclay), Andrea‌ Araldo (IMT), Nadjib Achir .
- Saeed Alsabbagh (UVSQ),‌ “Security of V2X Communications in 5g networks”, started‌ Sep. 2022, defended on Dec. 19, 2025. Advisors:‌ N. Aitsaadi, Cédric Adjih and A. Adouane.
- Felix Marcoccia (CIFRE) “Topology Optimization‌ in Mobile Wireless Networks‌ using Machine Learning”, started‌‌ 2022, defended on Oct. 13, 2025. Advisors: Paul‌ Mühlethaler and Cédric Adjih‌ . 43
- Najoua Benalya‌‌ (ENSI), “Agriculture de précision dans l'ère des drones‌ et d'intelligence artificielle”, started‌ in Nov. 2021, defended‌‌ on Dec 1st, 2025. Advisors: I. Amdouni, A.‌ Laouiti, L. Saidane, Cédric‌ Adjih .
- Khouloud Hwerbi‌‌ (ENSI), “Optimized Architectures and Algorithms for Blockchain and‌ IoT-based Applications”, started in‌ Nov. 2021, defended on‌‌ Nov. 24, 2025. Advisors: I. Amdouni, A. Laouiti,‌ L. Saidane, Cédric Adjih‌ .
PhDs supervision (in‌‌ progress):
- Mohramed Ourahou (CIFRE)"Transport on Demand (ToD) guided‌ by ML for maximizing‌ the accessibility of territories",‌‌ since Sep. 2024. Advisors: Aline Carneiro Viana ,‌ Andrea Araldo (TSP), Louis‌ Zigrand (PADAM Mobility).
- Wendlasida‌‌ Ouedraogo , “Vers l’exploitation des réseaux hétérogènes”, since‌ january 2024. Advisors: Nadjib‌ Achir , Lucas-Brehon Grataloup‌‌ (IMT), Antoine Lavignotte (IMT) and Andrea Araldo (IMT).‌
- Amira Dhaouadi , “Split‌ computing for constrained devices”,‌‌ since january 2024. Advisors: Cédric Adjih and Nadjib‌ Achir .
- Lucas Airam‌ Castro de Souza ,‌‌ “Anomaly Detection for Vehicular Networks”, since Nov. 2023.‌ Advisors: Miguel Elias Mitre‌ Campista, and Luís Henrique‌‌ Maciel Kosmalski Costa (GTA, UFRJ), and Nadjib Achir‌ . This Phd started‌ at the UFRJ and‌‌ co-supervision agreemed must be set up in 2024.‌
- Lucas Gabriel Da Silva‌ Felix (UFMG, Inria TRiBE),‌‌ “Assessing Shadows in Mobility: Beyond Spatiotemporal Patterns”, since‌ 2024. Advisors: Aline Carneiro‌ Viana, Anne Josiane Kouam,‌‌ Jussara Almeida (UFMG), Nadjib Achir.
- Haron C. Fantecele‌ (LNCC, Brazil), “Mathematical modeling‌ and machine learning applied‌‌ to human mobility prediction”, since Feb. 2020. Advisor:‌ Aline C. Viana and‌ Antonio Tadeu (LNCC).
- Corentin‌‌ Gautier (CIFRE) “FANET for Vehicle Swarms”, since 2022.‌ Advisors: Paul Mühlethaler .‌ Emmanuel Baccelli and Cédric‌‌ Adjih .
- Niruth Savin Bogahawatta , PhD at‌ University of Sydney, is‌ doing an 6-month internship‌‌ in our team, advised by Nadjib Achir ,‌ Aline Carneiro Viana ,‌ and Kanchana Thilakarathna (Univ.‌‌ of Sydney, Australia).
Master supervision:
- the team regularly‌ hosts master students and‌ PhD interns for periods‌‌ of 3 to 6 months. The list of‌ students/interns concerned by this‌ report year is mentioned‌‌ in team members list.
- Minxuan Wang (ENSTA amd‌ Shanghai Jiao Tong University)‌ advised by Cédric Adjih‌‌ , on “Semantic Communications for Digital Twin Systems‌ with Deep Learning”, in‌ 2025.
Bachelor’s supervision:
- Marta‌‌ Theodora Trales and Andrei-Valentin Stirbu advised by Cédric‌ Adjih , both students‌ from the Polytechnique Bachelor‌‌ program. They each presented their work at the‌ Junior Conference JWOC 2025,‌ and Andrei-Valentin Stirbu also‌‌ has an accepted article/presented at PEMWN 2025 42‌.
- Gustavo bruno dos‌ Santos , student from‌‌ UTFPR (Brazil) advised by Aline Carneiro Viana and‌ Thiago Silva (UTFPR) since‌ December 2025.

11.2.3 Juries‌‌

Aline Carneiro Viana : Selection committee: for Inria‌ Researchers with disabilities (CRCN-TH,‌ Chercheur/Chercheuse en situation de‌‌ Handicap); HDR juries: (as Reviewer) S. Kallel‌ Khemiri, ”Gestion et Contrôle‌ des Réseaux Véhiculaires de‌‌ l’IEEE 802.11p à la‌ 6G” (UVSQ-UPSaclay, Dec. 2025). PhD juries: (initially as‌ Examinateur then, for time constraint, as Invited) J.‌ XU, ”Characterisation of Anomalous Behaviour for Security in‌ Deep-Edge Wireless Systems” (FUN/Inria, Dec. 2025).
Cédric Adjih‌ : Selection committees: for Associate Professor at Université‌ Paris-Saclay.
Nadjib Achir : Selection committees: for Associate‌ Professor at Université de Technologie de Compiègne
Nadjib‌ Achir served as examiner for the PhD defense‌ of Najoua Benalya (ENSI), “Agriculture de précision dans‌ l'ère des drones et d'intelligence artificielle”, Dec 1st,‌ 2025.
Nadjib Achir served as examiner for the‌ PhD defense of Sekinat Oluwakunmi Yahya (INSA Lyon),‌ “A Study of Energy Consumption Challenges in Extended‌ Reality Services over Cellular Networks”, Jul. 24th, 2025.‌
Nadjib Achir served as examiner for the 2nd‌ year PhD evaluation of Zhaoxin CHANG, Telecom SudParis,‌ Institut Polytechnique de Paris, Oct. 2025.

11.3 Popularization‌

11.3.1 Scientific outreach and public engagement

Anne Josiane‌ Kouam : Panelist, WiSe 2025 Roundtable (Women‌ in Computer Science), December 2025. She contributed to‌ discussions on career paths, challenges, and opportunities in‌ computer science research for female students, sharing perspectives‌ on academic trajectories and support mechanisms for women‌ in computing.
Anne Josiane Kouam : Co-organizer,‌ Festival de Mathématiques, Yaoundé, Cameroon, August 2025. She‌ represented Animath International at a week-long mathematics festival‌ organized with Promo-Maths Cameroon and supported by Animath‌ France. She delivered outreach lectures to high-school students‌ and contributed to the Miss STEM Cameroon competition‌ promoting girls’ engagement in science.
Anne Josiane Kouam‌ : Panelist, Inauguration of the Just Do‌ Maths exposition, Inria Saclay, June 2025. She participated‌ in a public panel discussion on women’s careers‌ in mathematics and computer science, addressing researchers, educators,‌ and students.
Geoffrey Deperle participated of the program‌ “1 scientifique, 1 classe : Chiche!”, where‌ he presented research careers at Lycée Les 7‌ Mares (Maurepas), aiming to challenge stereotypes and highlight‌ the diversity of academic paths.
Geoffrey Deperle contributed‌ to an observation internship for seconde students by‌ introducing the TRiBE team and research activities, followed‌ by a hands-on session on Python-based fractal design‌ to illustrate geometric transformations and computational thinking.

11.3.2‌ Productions (articles, videos, podcasts, serious games, ...)

Aline‌ Carneiro Viana and Nadjib Achir were interviewed in‌ the article entitled ”Mobile networks: how can our‌ movements be tracked while preserving our anonymity?” in‌ the context of Smart Cities and Territories Inria‌ article series.
Aline Carneiro Viana was interviewed in‌ the article entitled ”NetMob 2025 in Paris: understanding‌ society through mobility data” in the context of‌ Networks Inria article series.

12 Scientific production

12.1‌ Major publications

1 articleS.Soumya Banerjee,‌ D.Dominique Barthel, Q.Quentin Lampin,‌ M.Marion Dumay, S.Stéphane Coutant,‌ C.Cédric Adjih, P.Paul Muhlethaler and‌ T.Thomas Watteyne. Automated Header Compression in‌ Constrained Networks.IEEE Communications Standards MagazineOctober‌ 2024. In press. HAL
2 inproceedingsS.‌Stefan Czybik, A. J.Anne Josiane Kouam, P.Peter Heubl‌, J. M.Jan‌ Magnus Nold and K.‌‌Konrad Rieck. A Large-Scale Study of Personalized‌ Phishing using Large Language‌ Models.USENIX 2026‌‌ - 35th USENIX Security SymposiumBaltimore (MD), United‌ StatesAugust 2026HAL‌
3 inproceedingsJ. P.‌‌João Paulo Esper, A. C.Aline Carneiro‌ Viana and J. M.‌Jussara M. Almeida.‌‌ Beauty or beast: human behavioral insights and learning‌ power of federated mobility‌ prediction.ACM SIGSPATIAL‌‌ 2024ACM SIGSPATIAL - 32nd International Conference on‌ Advances in Geographic Information‌ SystemsAtlanta, United States‌‌November 2024, 325-337HAL DOI
4 inproceedings‌I.Iman Hmedoush,‌ C.Cédric Adjih,‌‌ P.Paul Mühlethaler and L.Lou Salaun.‌ Multi-Power Irregular Repetition Slotted‌ ALOHA in Heterogeneous IoT‌‌ networks.PEMWN 2020 - 9th IFIP/IEEE International‌ Conference on Performance Evaluation‌ and Modeling in Wired‌‌ and Wireless NetworksPEMWN 2020 - 9th IFIP/IEEE‌ International Conference on Performance‌ Evaluation and Modeling in‌‌ Wired and Wireless Networks,Berlin / Virtual, Germany‌December 2020HAL
5‌ inproceedingsZ.Zhaolan Huang‌‌ and E.Emmanuel Baccelli. msf-CNN: Patch-based Multi-Stage‌ Fusion with Convolutional Neural‌ Networks for TinyML.‌‌NeurIPS 2025 - 39th Annual Conference on Neural‌ Information Processing SystemsSan‌ Diego, United StatesOctober‌‌ 2025HAL
6 articleP.Philippe Jacquet.‌ Information Theoretic Study of‌ Covid 19 Genome.‌‌Entropy263March 2024, 223HAL‌
7 inproceedings A. J.‌Anne Josiane Kouam,‌‌ A.Aline Carneiro Viana and A.Alain Tchana‌. Battle of Wits:‌ To What Extent Can‌‌ Fraudsters Disguise Their Tracks in International bypass Fraud?‌ ACM ASIACCS 2024 -‌ 19th ACM Asia Conference‌‌ on Computer and Communications Security Singapore, Singapore July‌ 2024 HAL DOI
8‌ inproceedingsA. J.Anne‌‌ Josiane Kouam, A. C.Aline Carneiro Viana‌, M. G.Mariano‌ G. Beiró, L.‌‌Leo Ferres and L.Luca Pappalardo. Beyond‌ Aggregates: A Fine-Grained Analysis‌ of Individual Mobility and‌‌ Traffic Dependencies.MSWiM 2025 - 27th International‌ Conference on Modeling, Analysis‌ and Simulation of Wireless‌‌ and Mobile SystemsBarcelona, SpainIEEEOctober 2025‌, 201-210HAL DOI‌
9 inproceedingsS.Shenghao‌‌ Yuan, F.Frédéric Besson, J.-P.Jean-Pierre‌ Talpin, S.Samuel‌ Hym, K.Koen‌‌ Zandberg and E.Emmanuel Baccelli. End-to-end Mechanized‌ Proof of an eBPF‌ Virtual Machine for Micro-controllers‌‌.CAV 2022 - 34th International Conference on‌ Computer Aided VerificationHaifa,‌ IsraelAugust 2022,‌‌ 1-23HAL
10 inproceedingsK.Koen Zandberg,‌ E.Emmanuel Baccelli,‌ S.Shenghao Yuan,‌‌ F.Frédéric Besson and J.-P.Jean-Pierre Talpin.‌ Femto-Containers: Lightweight Virtualization and‌ Fault Isolation For Small‌‌ Software Functions on Low-Power IoT Microcontrollers.Middleware‌ 2022 - 23rd ACM/IFIP‌ International Conference MiddlewareQuebec,‌‌ CanadaNovember 2022, 1-12HAL DOI

12.2‌ Publications of the year‌

International journals

11 article‌‌N.Najoua Benalaya, I.Ichrak Amdouni,‌ C.Cedric Adjih,‌ A.Anis Laouiti and‌‌ L. A.Leila Azouz‌ Saidane. Linear programming for UAVs search path‌ planning in livestock health monitoring.Computers and‌ Electronics in Agriculture240January 2026, 111232‌HAL DOI back to text
12 articleM.‌Michael Drmota, P.Philippe Jacquet, C.‌Changlong Wu and W.Wojciech Szpankowski. Precise‌ Regularized Minimax Regret with Unbounded Weights.IEEE‌ Transactions on Information Theory7111November 2025‌, 8753-8776HAL DOI
13 articleJ. P.‌João Paulo Esper, L. d.Luciano de‌ S. Fraga, A.Aline Carneiro Viana,‌ K. V.Kleber Vieira Cardoso and S. L.‌Sand Luz Correa. +Tour: Recommending personalized itineraries‌ for smart tourism.Computer Networks260April‌ 2025, 111118HALDOI back to text‌
14 articleK.Khouloud Hwerbi, I.Ichrak‌ Amdouni, C.Cédric Adjih, L. A.‌Leila Azouz Saidane and A.Anis Laouiti.‌ Delay Analysis of BFT Consensus : Case Study‌ of Narwhal and Bullshark Protocols.Computer Communications‌242October 2025, 108278HAL DOI back‌ to text
15 articleK.Khouloud Hwerbi,‌ I.Ichrak Amdouni, C.Cédric Adjih,‌ L. A.Leila Azouz Saidane and A.Anis‌ Laouiti. Delays in mempool-based blockchains under realistic‌ conditions: case of Narwhal and DroneVet application.‌Annals of Telecommunications - annales des télécommunicationsJanuary‌ 2026HAL DOI back to text
16 article‌R.Rosario Patanè, N.Nadjib Achir,‌ A.Andrea Araldo and L.Lila Boukhatem.‌ Vehicular cloud computing: a cost-effective alternative to edge‌ computing in 5G networks.Computer Networks268‌August 2025, 111365HAL DOI back to‌ text
17 articleP.Paulo Rettore, J.‌Jannis Mast, T.Thorsten Aurisch, A.‌ C.Aline Carneiro Viana, P.Peter Sevenich‌ and B.Bruno Santos. Military IoT from‌ Management to Perception: Challenges and Opportunities Across Layers‌.IEEE Internet of Things Magazine82‌March 2025, 25-31HAL DOI back to‌ text
18 articleS.Sanjeev Sharma, P.‌Praveen Singya, K.Kuntal Deka, C.‌Cedric Adjih and M.Mohit Sharma. Terahertz‌ Communication: State-of-the-Art and Future Directions.IEEE Open‌ Journal of the Communications Society62025,‌ 6281-6322HAL DOI back to text

International peer-reviewed‌ conferences

19 inproceedingsS.Saeed Alsabbagh, C.‌Cédric Adjih, A.Amine Adouane and N.‌Nadjib Aitsaadi. IRSA Under Capture Effect and‌ Imperfect SIC: A DE Analysis for Future Cellular‌ IoT.PIMRC 2025 - IEEE 36th International‌ Symposium on Personal, Indoor and Mobile Radio Communications‌Istanbul, TurkeyIEEESeptember 2025, 1-6HAL‌DOI back to text
20 inproceedingsS.Saeed‌ Alsabbagh, C.Cédric Adjih, A.Amine‌ Adouane and N.Nadjib Aitsaadi. Optimization of‌ Irregular Repetition Slotted ALOHA with Imperfect SIC in‌ 5G CIoT.ICC 2025 - IEEE International‌ Conference on CommunicationsMontreal, CanadaIEEEJune 2025‌, 5234-5240HAL DOIback to text
21 inproceedingsB.Bastien Buil‌, C.Chrystel Gaber‌, E.Emmanuel Baccelli‌‌ and S.Samia Bouzefrane. TinyML as a‌ Service on Multi-Tenant Microcontrollers‌.EWSN 2025 -‌‌ 22nd International Conference on Embedded Wireless Systems and‌ NetworksLeuven, BelgiumSeptember‌ 2025HAL back to‌‌ text
22 inproceedingsB.Bastien Buil, C.‌Chrystel Gaber, S.‌Sylvain Plessis, E.‌‌Emmanuel Baccelli and S.Samia Bouzefrane. Shared‌ Responsibility in Multi-Tenant Microcontrollers‌.CNSM 2025 -‌‌ 21st International Conference on Network and Service Management‌Bologna, ItalyOctober 2025‌HAL back to text‌‌
23 inproceedingsG.Geoffrey Deperle and P.Philippe‌ Jacquet. Generating Random‌ Hyperfractal Cities.Geometric‌‌ Science of Information7th International Conference, GSI 2025, Saint-Malo,‌ France, October 29–31, 2025,‌ Proceedings, Part IGSI‌‌ 2025 - 7th International Conference on Geometric Science‌ of InformationLecture Notes‌ in Computer ScienceLecture‌‌ Notes in Computer ScienceLNCS-16033Saint-malo, FranceSpringer‌ Nature SwitzerlandOctober 2026‌, 245-253HAL DOI‌‌back to text
24 inproceedingsA.Amira Dhaouadi‌, N.Nadjib Achir‌ and C.Cédric Adjih‌‌. Enhancing Split ViT Inference Through Sparsity-Driven Compression‌.VTC 2025 -‌ Vehicular Technology ConferenceOslo,‌‌ NorwayJune 2025HALback to text
25‌ inproceedingsC. S.Carlos‌ Sulbaran Fandino, A.‌‌ J.Anne Josiane Kouam and K.Konrad Rieck‌. Tiny Sensors, Big‌ Threats: Assessing Motion Sensor-based‌‌ Fingerprinting in Mobile Systems.MSWiM 2025 -‌ 27th International Conference on‌ Modeling, Analysis and Simulation‌‌ of Wireless and Mobile SystemsBarcelona, SpainIEEE‌October 2025, 286-295‌HAL DOI back to‌‌ text
26 inproceedingsL.Lucas Félix, A.‌ J.Anne Josiane Kouam‌, A.Aline Carneiro‌‌ Viana, N.Nadjib Achir and J. M.‌Jussara M Almeida.‌ Characterizing User Exposure in‌‌ Mobility Data: Insights from Varying User Behavioral Profiles‌.NetMob 2025 -‌ 9th edition of NetMob‌‌Paris, FranceOctober 2025HAL back to text‌back to text
27‌ inproceedingsL. G.Lucas‌‌ G S Félix, A. J.Anne Josiane‌ Kouam, A.Aline‌ Carneiro Viana, N.‌‌Nadjib Achir and J. M.Jussara M Almeida‌. The Silent Signature:‌ Behavior-based User Exposure in‌‌ Mobility Data.MDM 2025 - 26th IEEE‌ International Conference on Mobile‌ Data ManagementIrvine (CA),‌‌ United StatesJune 2025HAL DOI back to‌ text back to text‌
28 inproceedingsF.Frédéric‌‌ Fort, H.Hugo Forraz, K.Koen‌ Zandberg, G.Gilles‌ Grimaud and E.Emmanuel‌‌ Baccelli. Microcontroller Software Continuous Deployment: Secure Partial‌ Update Realms for RIOT‌ (PURR).DCOSS-IoT 2025‌‌ - 21st International Conference on Distributed Computing in‌ Smart Systems and the‌ Internet of ThingsLucca,‌‌ ItalyJune 2025HALback to text
29‌ inproceedingsE.Elena Frank‌, K.Kaspar Schleiser‌‌, R.Romain Fouquet, K.Koen Zandberg‌, C.Christian Amsüss‌ and E.Emmanuel Baccelli‌‌. Ariel OS: An Embedded Rust Operating System‌ for Networked Sensors &‌ Multi-Core Microcontrollers.DCOSS-IoT‌‌ 2025 - IEEE International‌ Conference on Distributed Computing in Smart Systems and‌ the Internet of ThingsLucca, ItalyIEEEJune‌ 2025HAL back to text
30 inproceedingsZ.‌Zhaolan Huang and E.Emmanuel Baccelli. msf-CNN:‌ Patch-based Multi-Stage Fusion with Convolutional Neural Networks for‌ TinyML.NeurIPS 2025 - 39th Annual Conference‌ on Neural Information Processing SystemsSan Diego, United‌ StatesOctober 2025HALback to text
31‌ inproceedingsA. J.Anne Josiane Kouam, A.‌Aline Carneiro Viana, P.Philippe Martins,‌ C.Cédric Adjih and A.Alain Tchana.‌ SigN: SIMBox Activity Detection Through Latency Anomalies at‌ the Cellular Edge.ACM Asia Conference on‌ Computer and Communications Security (ASIA CCS ’25)Hanoi,‌ VietnamAugust 2025HALDOI back to text‌back to text
32 inproceedingsA. J.Anne‌ Josiane Kouam, A. C.Aline Carneiro Viana‌, M. G.Mariano G. Beiró, L.‌Leo Ferres and L.Luca Pappalardo. Beyond‌ Aggregates: A Fine-Grained Analysis of Individual Mobility and‌ Traffic Dependencies.MSWiM 2025 - 27th International‌ Conference on Modeling, Analysis and Simulation of Wireless‌ and Mobile SystemsBarcelona, SpainIEEEOctober 2025‌, 201-210HAL DOIback to text back‌ to text
33 inproceedingsF.Fernando Molano Ortiz‌, G.Guillaume Farhi-Rivasseau, N.Nadjib Achir‌ and A.Aline Carneiro Viana. (Demo) MITIK‌ Toolkit: A Privacy-Compliant Passive Collection of WiFi Probe‌ Request Datasets.ACM SIGCOMM Posters and Demos‌ '25: Proceedings of the ACM SIGCOMM 2025 Posters‌ and DemosACM SIGCOMM 2025 Posters and Demos‌Coimbra Portugal, PortugalACMSeptember 2025, 193-195‌HAL DOI back to text
34 inproceedingsT.-S.‌Thanh-Son-Lam Nguyen, N.Nadjib Aitsaadi and C.‌Cedric Adjih. ANSB : An Optimized Network‌ Slicing Scheme for Adaptive Load Balancing in 5G‌ Core Network.ICC 2025 - IEEE International‌ Conference on CommunicationsMontréal, Canada2025, 1402-1407‌HAL DOI back to text
35 inproceedingsT.‌Tonmoy Rajkhowa, A.Amit Singh, S.‌Sanjeev Sharma, K.Kuntal Deka and C.‌Cedric Adjih. DSCAN-Net: A Dual-Stream Network for‌ Classifying Modulation Schemes in OTFS Systems.NCC‌ 2025 - 31st edition of the National Conference‌ on CommunicationsNew Delhi, IndiaIEEEMarch 2025‌, 1-6HAL DOIback to text
36‌ inproceedingsG. M.Gaelle Manuela Yonga, A.‌ J.Anne Josiane Kouam, A.Aline Carneiro‌ Viana and A. V.Auguste Vigny Noumsi Woguia‌. Évaluation de l'utilisabilité et de la fiabilité‌ de données spatio-temporelles anonymisées.CORES 2025 -‌ 10èmes Rencontres Francophones sur la Conception de Protocoles,‌ l'Evaluation de Performances et l'Expérimentation des Réseaux de‌ CommunicationCORES 2025 - 10èmes Rencontres Francophones sur‌ la Conception de Protocoles, l'Evaluation de Performances et‌ l'Expérimentation des Réseaux de CommunicationSaint Valery-sur-Somme, France‌June 2025HAL back to text
37 inproceedings‌G.Gaelle Yonga, A. J.Anne Josiane‌ Kouam, A. C.Aline Carneiro Viana and‌ A.Auguste Noumsi. On Assessing Usability and Reliability of Anonymized Spatio-Temporal‌ Data.UrbCom -‌ 7th International Workshop on‌‌ Urban Computing, Co-located with IEEE DCOSS-IoT 2025Lucca,‌ ItalyIEEEJune 2025‌, 689-696HAL DOI‌‌back to text

Conferences without proceedings

38 inproceedings‌L. A.Lucas Airam‌ Castro de Souza,‌‌ M.Matteo Sammarco, N.Nadjib Achir,‌ L. H.Luís Henrique‌ Maciel Kosmalski Costa and‌‌ M. E.Miguel Elias Mitre Campista. TOFL:‌ Time Optimized Federated Learning‌.SBSeg 2025 -‌‌ Simpósio Brasileiro de CibersegurançaFoz Do Iguaçu, Brazil‌September 2025HAL back‌ to text
39 inproceedings‌‌S.Stefan Czybik, A. J.Anne Josiane‌ Kouam, P.Peter‌ Heubl, J. M.‌‌Jan Magnus Nold and K.Konrad Rieck.‌ A Large-Scale Study of‌ Personalized Phishing using Large‌‌ Language Models.USENIX 2026 - 35th USENIX‌ Security SymposiumBaltimore (MD),‌ United StatesAugust 2026‌‌HAL
40 inproceedingsL. G.Lucas G S‌ Félix, N.Nadjib‌ Achir, A. J.‌‌Anne Josiane Kouam, A.Aline Carneiro Viana‌ and J. M.Jussara‌ Marques Almeida. Measuring‌‌ the users exposure in mobility data.SBRC‌ 2025 - 43° Simpósio‌ Brasileiro de Redes de‌‌ Computadores e Sistemas DistribuídosNatal - Rio Grande‌ do Norte, BrazilMay‌ 2025HAL DOI back‌‌ to text back to text
41 inproceedingsM.‌Mohamed Ourahou, A.‌Andrea Araldo, L.‌‌Louis Zigrand and A.Aline Carneiro Viana.‌ Towards assessing accessibility resulting‌ from integrating demand-responsive and‌‌ conventional public transport with travel time uncertainty.‌NetMob 2025Paris CNAM,‌ FranceOctober 2025HAL‌‌back to text
42 inproceedingsA.-V.Andrei-Valentin Stirbu‌ and C.Cédric Adjih‌. FIT-IRSA: Feedback-Integrated Two-Phase‌‌ IRSA with Deep Reinforcement Learning.PEMWN 2025‌ - 14th IFIP/IEEE International‌ Conference on Performance Evaluation‌‌ and Modeling in Wired and Wireless NetworksParis,‌ FranceIEEENovember 2025‌, 1-10HAL DOI‌‌back to text back to text

Doctoral dissertations‌ and habilitation theses

43‌ thesisF.Félix Marcoccia‌‌. Topology optimization in mobile wireless networks using‌ machine learning.Sorbonne‌ UniversitéOctober 2025HAL‌‌back to text back to text
44 thesis‌R.Rosario Patanè.‌ Technical, energy performance, economic‌‌ and environmental evaluation of vehicular and edge computing‌.Université Paris-SaclayNovember‌ 2025HAL back to‌‌ text

Reports & preprints

45 reportN.Nadjib‌ Achir, A.Aline‌ Carneiro Viana, M.‌‌Marcelo Dias de Amorim, A.Anne Fladenmuller‌, Y.Yacine Ghamri-Doudane‌ and J.-L.Jean-Loup Guillaume‌‌. D5.3: Exploitation plan.D5.3Sorbonne Université‌ (Paris, France); Inria Saclay‌ - Île de France;‌‌ La Rochelle Université, FranceFebruary 2025HAL
46‌ miscN.Nadjib Achir‌ and P.Philippe Jacquet‌‌. BUBBLE-BLUE a multihop private network based on‌ Bluetooth.January 2026‌HAL DOI back to‌‌ text
47 miscL. A.Lucas Airam Castro‌ de Souza, M.‌Matteo Sammarco, N.‌‌Nadjib Achir, M. E.Miguel Elias Mitre‌ Campista and L. H.‌Luís Henrique Maciel Kosmalski‌‌ Costa. TOFL: Time‌ Optimized Federated Learning.March 2025HAL
48‌ miscA.Alexandre Chasse, A. J.Anne‌ Josiane Kouam, A.Aline Carneiro Viana,‌ R.Razvan Stanica, W.Wellington Lobato,‌ G.Geymerson Ramos, G.Geoffrey Deperle,‌ A.Abdelmounaim Bouroudi, S.Suzanne Bussod and‌ F.Fernando Molano Ortiz. The NetMob25 Dataset:‌ A High-resolution Multi-layered View of Individual Mobility in‌ Greater Paris Region.2025HAL
49 report‌G.Guillaume Farhi-Rivasseau, F.Fernando Molano Ortiz‌, A.Aline Carneiro Viana and N.Nadjib‌ Achir. MITIK-GUIDE.Inria Saclay - Île‌ de FranceJanuary 2025HAL
50 miscA.‌ J.Anne Josiane Kouam, A.Aline Carneiro‌ Viana, M.Mariano Beiró, L.Leo‌ Ferres and L.Luca Pappalardo. Characterizing User‌ Behavior: The Interplay Between Mobility Patterns and Mobile‌ Traffic.2025HALDOI
51 miscF.‌Félix Marcoccia, V.Victor Fagoo, G.‌Gilles Monzat, C.Cédric Adjih, T.‌Thomas Watteyne and P.Paul Mühlethaler. NetDiff:‌ Graph Diffusion with Improved Global Capabilities to Generate‌ and Update Mobile Network Topologies.December 2025‌HAL
52 miscF.Félix Marcoccia, V.‌Victor Fagoo, G.Gilles Monzat de Saint‌ Julien, C.Cédric Adjih, T.Thomas‌ Watteyne and P.Paul Mühlethaler. TopoFormer: An‌ Efficient Link-Set Prediction Architecture for Ad Hoc Network‌ Topology Generation.September 2025HAL back to‌ text

Other scientific publications

53 miscN.Nicolas‌ Cassiau, N.Nadjib Achir, C.Cédric‌ Adjih, G.Guillaume Andrieux, W.Walid‌ Bechkit, S.Siwar Ben Hadj Said,‌ O.Olivier Boissier, A.Aymen Bouferroum,‌ R.Richard Combes, F.Fabien Courrèges,‌ H.Hiba Dakdouk, A.Amira Dhaouadi,‌ J.-F.Jean-François Diouris, S. E.Salah Eddine‌ Elayoubi, J.Jérôme Härri, M.Mihia‌ Kassi, X.Xavier Lagrange, Y.Yandi‌ Liu, V.Valeria Loscri, V.Valerian‌ Mannoni, S.Sébastien Maudet, N.Nathalie‌ Mitton, A.Amina Mokdad, E.Emmanuel‌ Moulay, A.Ali Nadar, R.Rémi‌ Nahon, R.Romain Négrier, v. T.‌van Tam Nguyen, A.Alexander Pelov,‌ C.Clency Perrine, M.Marc Pierre,‌ S.Sébastien Pillement, A.Antony Pottier,‌ C.Charly Poulliat, Y.Yannis Pousset,‌ M.Mina Rady, N.Nourhen Sboui,‌ P.Patrick Sondi and L.Laurent Toutain.‌ Overcoming the Technical Hurdles of IoT Adoption: the‌ FITNESS Project Vision and Insights.September 2025‌HAL DOI
54 miscL.Lucas Félix,‌ A. J.Anne Josiane Kouam, A.Aline‌ Carneiro Viana, N.Nadjib Achir and J.‌ M.Jussara M Almeida. Investigating different perspectives‌ on exposure in mobility datasets.June 2025‌HAL

Software

55 softwareA. J.Anne Josiane‌ Kouam, A.Aline Carneiro Viana and A.‌Alain Tchana. En-WDM.1.0December 2025 lic: GNU General Public‌ License v3.0 or later‌.HAL Software Heritage‌‌VCS
56 softwareA. J.Anne Josiane Kouam‌, A.Aline Carneiro‌ Viana and A.Alain‌‌ Tchana. FraudZen.1.0December 2025 lic:‌ GNU General Public License‌ v3.0 or later.‌‌HAL Software Heritage VCS

12.3 Cited publications

57‌ miscT. E.The‌ Edge AI & Vision‌‌ Alliance. The microcontroller market shifting trends and‌ price surges.[Online;‌ accessed 14-Nov-2024]2023back‌‌ to text
58 articleL.Licia Amichi,‌ A.Aline Carneiro Viana‌, M.Mark Crovella‌‌ and A. A.Antonio A. F Loureiro.‌ Revealing an inherently limiting‌ factor in human mobility‌‌ prediction.IEEE Transactions on Emerging Topics in‌ Computing2022HAL DOI‌back to text
59‌‌ articleB.Bartlomiej Blaszczyszyn, P.Philippe Jacquet‌, B.Bernard Mans‌ and D.Dalia Popescu‌‌. Energy and Delay Trade-Offs of End-to-End Vehicular‌ Communications using a Hyperfractal‌ Urban Modelling.Annals‌‌ of Telecommunications - annales des télécommunications2023HAL‌DOI back to text‌
60 articleP.Pedro‌‌ Cruz, N.Nadjib Achir and A.Aline‌ Carneiro Viana. On‌ the Edge of the‌‌ Deployment: A Survey on Multi-Access Edge Computing.‌ACM Computing Surveys55‌52022, 1-34‌‌HAL DOI back to text
61 miscQ.‌Qiang Duan, S.‌Shijing Hu, R.‌‌Ruijun Deng and Z.Zhihui Lu. Combined‌ Federated and Split Learning‌ in Edge Computing for‌‌ Ubiquitous Intelligence in Internet of Things: State of‌ the Art and Future‌ Directions.2022,‌‌ URL: https://arxiv.org/abs/2207.09611back to text
62 reportH.‌ C.Haron C Fantecele‌, S.Solohaja Rabenjamina‌‌, A.Aline Carneiro Viana, R.Razvan‌ Stanica and A.Artur‌ Ziviani. SafeCityMap (1st‌‌ phase) -- COVID INRIA mission: Investigating population mobility‌ habits in metropolitan zones‌ and the lockdown impact‌‌ using mobile phone data.InriaMay 2021‌HAL back to text‌
63 bookP.Philippe‌‌ Jacquet. Paradoxes and Physical Limits of Information‌ Theory.1World‌ Scientific Series on Quantum‌‌ Algorithms, Information, and LearningWorld Scientific2025,‌ 336DOI back to‌ text
64 inproceedingsA.‌‌Abhishek Kumar Mishra, A.Aline Carneiro Viana‌, N.Nadjib Achir‌ and C.Catuscia Palamidessi‌‌. Public Wireless Packets Anonymously Hurt You.‌IEEE LCN 2021 (Doctoral-track‌ - Promising ideas)Edmonton‌‌ / Virtual, CanadaOctober 2021HAL DOI back‌ to text
65 article‌E.Emanuel Lima,‌‌ A.Ana Aguiar, P.Paulo Carvalho and‌ A.Aline Carneiro Viana‌. Human Mobility Support‌‌ for Personalised Data Offloading.IEEE Transactions on‌ Network and Service Management‌192February 2022‌‌, 1505-1520HAL DOIback to text
66‌ phdthesisA. K.Abhishek‌ Kumar Mishra. Revealing‌‌ and exploiting privacy vulnerabilities in users' public wireless‌ packets.Institut Polytechnique‌ de ParisOctober 2023‌‌HAL back to text
67 articleA. C.‌Ana Claudia B. L.‌ Monção, S. L.‌‌Sand Luz Correa,‌ A.Aline Carneiro Viana and K. V.Kleber‌ Vieira Cardoso. Combining Resource-Aware Recommendation and Caching‌ in the Era of MEC for Improving the‌ Experience of Video Streaming Users.IEEE Transactions‌ on Services ComputingOctober 2022, 1-14HAL‌DOI back to text
68 articleD.Dalia‌ Popescu, P.Philippe Jacquet and B.Bernard‌ Mans. Connecting flying backhauls of unmanned aerial‌ vehicles to enhance vehicular networks with fixed 5G‌ NR infrastructure.IET Smart Cities4September‌ 2022, 239 - 254HAL DOI back‌ to text
69 miscB.Bruce Schneier.‌ Click Here to Kill Everyone.NY Magazine,‌ 2017. [Online; accessed 14-Nov-2024]2017back to text‌
70 miscVentureBeat. Why TinyML is a‌ giant opportunity.[Online; accessed 14-Nov-2024]2020back‌ to text

TRIBE - 2025

TRIBE - 2025

2025﻿​﻿﻿Activity reportProject-TeamTRIBE​‌﻿﻿

Keywords

Computer Science﻿‌​‌ and Digital Science

Other﻿​​﻿ Research Topics and Application​​​‌ Domains

1 Team﻿​​﻿ members, visitors, external collaborators​​​‌

Research Scientists

Post-Doctoral Fellow

PhD Students

Technical Staff

Interns and Apprentices

Administrative Assistant

Visiting Scientist

External﻿​﻿﻿ Collaborators

2 Overall​‌﻿﻿ objectives

2.1 Projections and emerging​​​‌ challenges

2.2 Team vision and​‌﻿﻿ approach

3​​​‌ Research program

4 Application​​​‌ domains

5 Social and​​​‌ environmental responsibility

5.1 Footprint﻿﻿﻿‌ of research activities

5.2 Impact of research​‌﻿﻿ results

6 Highlights of the​‌﻿﻿ year

6.1 Awards﻿​﻿﻿

6.2 Scientific​‌﻿﻿ Events and Conferences

6.3 Books﻿​﻿﻿

7﻿﻿﻿‌ Latest software developments, platforms,﻿‌​‌ open data

7.1 Latest software developments﻿‌​‌

7.1.1 RIOT

7.1.2 FraudZen

7.1.3​​​‌ MITIK-MGMT

7.1.4​​​‌ MITIK-SENS

7.1.5 MITIK-LINK﻿​﻿﻿

7.1.6​​﻿﻿ MITIK-TRAJ

7.1.7 MobilityPulse

7.1.8 En-WDM

7.1.9 Ariel OS

7.1.10 RIOT-rs

7.1.11​​﻿﻿ MoBES

7.1.12﻿﻿﻿‌ ZenPlus

7.1.13​‌﻿﻿ MITIK-HAND

7.1.14 HEXPOSE

7.2 New﻿‌​‌ platforms

7.3 Open data​​﻿﻿

SigN Dataset: SIMBox Activity​​​‌ Detection Through Cellular Latency﻿​﻿﻿ Anomalies

FraudZen Dataset: Realistic Ground​‌﻿﻿ Truth CDRs of Bypass​​﻿﻿ Fraud Techniques

MITIK Dataset of pseudo-anonymized﻿﻿﻿‌ WiFi management frames from﻿‌​‌ a parking area

MITIK​​​‌ Dataset of WiFi pseudo-anonymised﻿﻿﻿‌ public management frames captured﻿‌​‌ at La Rochelle University﻿​​﻿

8 New results

8.1 [Axis 1]: Optimized﻿​​﻿ communication protocols

8.1.1 Modern​​​‌ Random Access: Irregular Repetition﻿﻿﻿‌ Slotted Aloha (IRSA)

8.1.2 An​‌﻿﻿ Introduction to Modern Random​​﻿﻿ Access Protocols for IoT​​​‌ Communications (book chapter)

8.1.3 Optimization﻿​﻿﻿ of Irregular Repetition Slotted​‌﻿﻿ ALOHA with Imperfect SIC​​﻿﻿ in 5G CIoT

8.1.4 IRSA​​​‌ Under Capture Effect and﻿﻿﻿‌ Imperfect SIC: a de﻿‌​‌ Analysis for Future Cellular﻿​​﻿ IoT

8.1.5 FIT-IRSA: Feedback-Integrated Two-Phase﻿﻿﻿‌ IRSA with Deep Reinforcement﻿‌​‌ Learning

8.1.6 Terahertz Communication: State-of-the-Art﻿​​﻿ and Future Directions

8.2​​​‌ [Axis 1] Tailored embedded﻿​﻿﻿ software platforms and Unified​‌﻿﻿ low-end IoT technology

8.2.1​​﻿﻿ Embedded Rust Operating System​​​‌ for Networked Sensors and﻿​﻿﻿ Multi-Core Microcontrollers

8.2.2 SLAs for Shared﻿​﻿﻿ Responsibility in Multi-Tenant Microcontrollers​‌﻿﻿

8.3 [Axis 1] Low-footprint﻿﻿﻿‌ cybersecurity mechanisms

8.3.1 Secure﻿‌​‌ Low-Power Software Continuous Deployment﻿​​﻿

8.3.2 Standardization of﻿﻿﻿‌ Secure Software Updates for﻿‌​‌ Low-power IoT Devices

8.4 [Axis﻿﻿﻿‌ 1] Edge AI and﻿‌​‌ Tiny Machine Learning

8.4.1﻿​​﻿ TinyML as a Service​​​‌ on Multi-Tenant Microcontrollers

8.4.2﻿​﻿﻿ RAM Footprint Optimizations for​‌﻿﻿ TinyML Inference with Convolutional​​﻿﻿ Neural Networks

8.5 [Axis﻿​﻿﻿ 2] Extracting high-end IoT​‌﻿﻿ footprints in data.

8.5.1​​﻿﻿ MITIK Toolkit: A Privacy-Compliant​​​‌ Passive Collection of WiFi﻿​﻿﻿ Probe Request Datasets [Axis​‌﻿﻿ 2]

8.5.2 Towards﻿​﻿﻿ assessing accessibility resulting from​‌﻿﻿ integrating demand-responsive and conventional​​﻿﻿ public transport with travel​​​‌ time uncertainty

8.6 [Axis 2] User﻿﻿﻿‌ pattern understanding at the﻿‌​‌ Internet edge.

8.6.1 Generating﻿​​﻿ Random Hyperfractal Cities

8.6.2﻿‌​‌ Predicting Mobility with Small﻿​​﻿ Data and Physical Principles​​​‌

8.6.3 A Fine-Grained Analysis​​​‌ of Individual Mobility and﻿​﻿﻿ Traffic Dependencies

8.6.4 Assessing﻿​﻿﻿ Usability and Reliability of​‌﻿﻿ Anonymized Spatio-Temporal Data

8.7﻿﻿﻿‌ [Axis 2] Addressing end-users﻿‌​‌ privacy exposure and security﻿​​﻿ concerns in networking data.​​​‌

8.7.1 Behavior-based User Exposure﻿﻿﻿‌ in Mobility Data

8.7.2 Sensor-based fingerprinting and​‌﻿﻿ privacy risks in mobile​​﻿﻿ systems

8.7.3﻿​﻿﻿ SigN: Empirical Insights and​‌﻿﻿ Practical Solution for SIMBox​​﻿﻿ Fraud Prevention at the​​​‌ Cellular Edge

8.8 [Axis 3]​​​‌ Decentralized network mechanisms and﻿﻿﻿‌ architectures

8.8.1 ANSB: An﻿‌​‌ Optimized Network Slicing Scheme﻿​​﻿ for Adaptive Load Balancing​​​‌ in 5G Core Network﻿﻿﻿‌

2025Activity reportProject-TeamTRIBE‌

Computer Science‌‌ and Digital Science

Other Research Topics and Application‌ Domains

1 Team members, visitors, external collaborators‌

External Collaborators

2 Overall‌ objectives

2.1 Projections and emerging‌ challenges

2.2 Team vision and‌ approach

3‌ Research program

4 Application‌ domains

5 Social and‌ environmental responsibility

5.1 Footprint‌ of research activities

5.2 Impact of research‌ results

6 Highlights of the‌ year

6.1 Awards

6.2 Scientific‌ Events and Conferences

6.3 Books

7‌ Latest software developments, platforms,‌‌ open data

7.1 Latest software developments‌‌

7.1.3‌ MITIK-MGMT

7.1.4‌ MITIK-SENS

7.1.5 MITIK-LINK

7.1.6 MITIK-TRAJ

7.1.11 MoBES

7.1.12‌ ZenPlus

7.1.13‌ MITIK-HAND

7.2 New‌‌ platforms

7.3 Open data

SigN Dataset: SIMBox Activity‌ Detection Through Cellular Latency Anomalies

FraudZen Dataset: Realistic Ground‌ Truth CDRs of Bypass Fraud Techniques

MITIK Dataset of pseudo-anonymized‌ WiFi management frames from‌‌ a parking area

MITIK‌ Dataset of WiFi pseudo-anonymised‌ public management frames captured‌‌ at La Rochelle University

8.1 [Axis 1]: Optimized communication protocols

8.1.1 Modern‌ Random Access: Irregular Repetition‌ Slotted Aloha (IRSA)

8.1.2 An‌ Introduction to Modern Random Access Protocols for IoT‌ Communications (book chapter)

8.1.3 Optimization of Irregular Repetition Slotted‌ ALOHA with Imperfect SIC in 5G CIoT

8.1.4 IRSA‌ Under Capture Effect and‌ Imperfect SIC: a de‌‌ Analysis for Future Cellular IoT

8.1.5 FIT-IRSA: Feedback-Integrated Two-Phase‌ IRSA with Deep Reinforcement‌‌ Learning

8.1.6 Terahertz Communication: State-of-the-Art and Future Directions

8.2‌ [Axis 1] Tailored embedded software platforms and Unified‌ low-end IoT technology

8.2.1 Embedded Rust Operating System‌ for Networked Sensors and Multi-Core Microcontrollers

8.2.2 SLAs for Shared Responsibility in Multi-Tenant Microcontrollers‌

8.3 [Axis 1] Low-footprint‌ cybersecurity mechanisms

8.3.1 Secure‌‌ Low-Power Software Continuous Deployment

8.3.2 Standardization of‌ Secure Software Updates for‌‌ Low-power IoT Devices

8.4 [Axis‌ 1] Edge AI and‌‌ Tiny Machine Learning

8.4.1 TinyML as a Service‌ on Multi-Tenant Microcontrollers

8.4.2 RAM Footprint Optimizations for‌ TinyML Inference with Convolutional Neural Networks

8.5 [Axis 2] Extracting high-end IoT‌ footprints in data.

8.5.1 MITIK Toolkit: A Privacy-Compliant‌ Passive Collection of WiFi Probe Request Datasets [Axis‌ 2]

8.5.2 Towards assessing accessibility resulting from‌ integrating demand-responsive and conventional public transport with travel‌ time uncertainty

8.6 [Axis 2] User‌ pattern understanding at the‌‌ Internet edge.

8.6.1 Generating Random Hyperfractal Cities

8.6.2‌‌ Predicting Mobility with Small Data and Physical Principles‌

8.6.3 A Fine-Grained Analysis‌ of Individual Mobility and Traffic Dependencies

8.6.4 Assessing Usability and Reliability of‌ Anonymized Spatio-Temporal Data

8.7‌ [Axis 2] Addressing end-users‌‌ privacy exposure and security concerns in networking data.‌

8.7.1 Behavior-based User Exposure‌ in Mobility Data

8.7.2 Sensor-based fingerprinting and‌ privacy risks in mobile systems

8.7.3 SigN: Empirical Insights and‌ Practical Solution for SIMBox Fraud Prevention at the‌ Cellular Edge

8.8 [Axis 3]‌ Decentralized network mechanisms and‌ architectures

8.8.1 ANSB: An‌‌ Optimized Network Slicing Scheme for Adaptive Load Balancing‌ in 5G Core Network‌

8.8.2 +Tour: Recommending personalized itineraries for smart tourism‌

8.8.3 BUBBLE-BLUE a‌ multihop private network based on Bluetooth

8.9 [Axis‌ 3] Machine Learning enhanced network protocols and classical‌ network optimization (methods and techniques)

8.9.1 Precise Regularized‌ Minimax Regret with Unbounded Weights

8.9.2 Time- and Latency-Aware‌‌ Machine Learning for Vehicular Networks: Fast Federated Training‌ and Device-Adaptive Model Selection‌

8.9.3 Topology optimization in‌ mobile wireless networks using‌‌ machine learning

8.9.4 TopoFormer:‌ An Efficient Link-Set Prediction Architecture for Ad Hoc‌ Network Topology Generation

8.9.5 DSCAN-Net: A Dual-Stream‌‌ Network for Classifying Modulation Schemes in OTFS Systems‌

8.9.6 Linear programming for UAVs search path planning‌ in livestock health monitoring‌

8.10‌ [Axis 3] Edge network offloading (methods and techniques)‌

8.10.1 Vehicular Cloud Computing as a Cost-Effective, Profitable,‌ and Sustainable Alternative to 5G Edge Computing

8.10.3 Enhancing‌ Split ViT Inference Through‌‌ Sparsity-Driven Compression

8.11 [Axis‌ 3] Security of the‌‌ edge/core compound including IoT deployments (technologies)

8.11.1 Delay‌ analysis of BFT consensus‌ : Case study of‌‌ Narwhal and Bullshark protocols

8.11.2 Delays in mempool-based blockchains under realistic‌ conditions: case of Narwhal and DroneVet application

8.11.3 Military IoT‌‌ from Management to Perception: Challenges and Opportunities Across‌ Layers

9 Bilateral contracts and‌ grants with industry

9.1‌‌ Bilateral grants with industry