Participants: Muhammad Jawad Khokhar, Thibaut Ehlinger, Chadi Barakat.  

Video streaming is a dominant contributor to the global Internet traffic. Consequently, monitoring video streaming Quality of Experience (QoE) is of paramount importance to network providers. Monitoring QoE of video is a challenge as most of the video traffic of today is encrypted. In this work, we consider this challenge and present an approach based on controlled experimentation and machine learning to estimate QoE from encrypted video traces using network level measurements only. We consider a case of YouTube and play out a wide range of videos under realistic network conditions to build ML models (classification and regression) that predict the subjective MOS (Mean Opinion Score) based on the ITU P.1203 model along with the QoE metrics of startup delay, quality (spatial resolution) of playout and quality variations, and this is using only the underlying network Quality of Service (QoS) features. We comprehensively evaluate our approach with different sets of input network features and output QoE metrics. Overall, our classification models predict the QoE metrics and the ITU MOS with an accuracy of 63-90% while the regression models show low error; the ITU MOS (1-5) and the startup delay (in seconds) are predicted with a root mean square error of 0.33 and 2.66 respectively. The results of this work were published in  and can be found with further details in the PhD manuscript of Muhammad Jawad Khokhar graduated in October 2019.

Participants: Imane Taibi, Chadi Barakat.  

Web browsing remains one of the dominant applications of the internet, so inferring network performance becomes crucial for both users and providers (access and content) so as to be able to identify the root cause of any service degradation. Recent works have proposed several network troubleshooting tools, e.g, NDT, MobiPerf, SpeedTest, Fathom. Yet, these tools are either computationally expensive, less generic or greedy in terms of data consumption. The main purpose of this work funded by the IPL BetterNet is to leverage passive measurements freely available in the browser and machine learning techniques (ML) to infer network performance (e.g., delay, bandwidth and loss rate) without the addition of new measurement overhead. To enable this inference , we propose a framework based on extensive controlled experiments where network configurations are artificially varied and the Web is browsed, then ML is applied to build models that estimate the underlying network performance. In particular, we contrast classical ML techniques (such as random forest) to deep learning models trained using fully connected neural networks and convolutional neural networks (CNN). Results of our experiments show that neural networks have a higher accuracy compared to classical ML approaches. Furthermore, the model accuracy improves considerably using CNN. These results were published in .

Participants: Othmane Belmoukadam, Muhammad Jawad Khokhar, Chadi Barakat.  

Screen resolution along with network conditions are main objective factors impacting the user experience, in particular for video streaming applications. Terminals on their side feature more and more advanced characteristics resulting in different network requirements for good visual experience. Previous studies tried to link MOS (Mean Opinion Score) to video bit rate for different screen types (e.g., CIF, QCIF, and HD). We leverage such studies and formulate a QoE-driven resource allocation problem to pinpoint the optimal bandwidth allocation that maximizes the QoE (Quality of Experience) over all users of a provider located behind the same bottleneck link, while accounting for the characteristics of the screens they use for video playout. For our optimization problem, QoE functions are built using curve fitting on data sets capturing the relationship between MOS, screen characteristics, and bandwidth requirements. We propose a simple heuristic based on Lagrangian relaxation and KKT (Karush Kuhn Tucker) conditions for a subset of constraints. Numerical simulations show that the proposed heuristic is able to increase overall QoE up to 20% compared to an allocation with TCP look-alike strategies implementing max-min fairness. Later, we use a MPEG/DASH implementation in the context of ns-3 and show that coupling our approach with a rate adaptation algorithm can help increasing QoE while reducing both resolution switches and number of interruptions. Our framework and the first validation results were published in .

Participants: Karyna Gogunska, Chadi Barakat.  

With the increasing popularity of cloud networking and the widespread usage of virtualization as a way to offer flexible and virtual network and computing resources, it becomes more and more complex to monitor this new virtual environment. Yet, monitoring remains crucial for network troubleshooting and analysis. Controlling the measurement footprint in the virtual network is one of the main priorities in the process of monitoring as resources are shared between the compute nodes of tenants and the measurement process itself. In this paper, first, we assess the capability of machine learning to predict measurement impact on the ongoing traffic between virtual machines; second, we propose a data-driven solution that is able to provide optimal monitoring parameters for virtual network measurement with minimum traffic interference. These results were published in  and are part of the PhD manuscript of Karyna Gogunska graduated in December 2019.

Participants: Houssam Elbouanani, Chadi Barakat.  

Data center network monitoring can be carried out at hardware networking equipment (e.g. physical routers) and/or software networking equipment (e.g. virtual switches). While software switches offer high flexibility to deploy various monitoring tools, they have to utilize server resources, esp. CPU and memory, that can no longer be reserved fully to service users' traffic. In this work we closely examine the costs of (i) sampling packets ; (ii) sending them to a user-space program for measurement; and (iii) forwarding them to a remote server where they will be processed in case of lack of resources locally. Starting from empirical observations, we derive an analytical model to accurately predict ($R^2=$ 99.5%) the three aforementioned costs, as a function of the sampling rate. We next introduce a collaborative approach for traffic monitoring and sampling that maximizes the amount of collected traffic without impacting the data center's operation. We analyze, through numerical simulations, the performance of our collaborative solution. The results show that it is able to take advantage of the uneven loads on the servers to maximize the amount of traffic that can be sampled at the scale of a data center. The resulting gain can reach 200% compared to a non collaborative approach. These results were published in .

Participants: Damien Saucez, Yevhenii Semenko.  

The sharing economy fundamentally changed business and social interactions. Interestingly, while in essence this form of collaborative economy allows people to directly interact with each other, it is also at the source of the advent of eminently centralized platforms and marketplaces, such as Uber and Airbnb. One may be concerned with the risk of giving the control of a market to a handful of actors that may unilaterally fix their own rules and threaten privacy. Within the Data Privacy project of the UCAJedi Idex Academy 5 and House of Human and Social Sciences, Technologies and Uses Theme, we have proposed a holistic solution to address privacy issues in the sharing economy. We considered the case of ridesharing and proposed a decentralized architecture which gives the opportunity to shift from centralized platforms to decentralized ones. Digital communications in our proposition are specifically designed to preserve data privacy and avoid any form of centralization. A blockchain is used in our proposition to guarantee the essential roles of a marketplace, but in a decentralized way. Our evaluation shows that privacy protection without trusted entities comes at the cost of harder scalability than an approach with a trusted third party. However, our numerical evaluation on real data and our Android prototype shows the practical feasibility of our approach. The results obtained in this activity are published in 12th International Conference on Security, Privacy, and Anonymity in Computation, Communication, and Storage (SpaCCS) 2019, Atlanta  and documented in a research report .

Participants: Imane Fouad, Arnaud Legout, Natasa Sarafijanovic-Djukic.  

Web tracking has been extensively studied over the last decade. To detect tracking, previous studies and user tools rely on filter lists. However, it has been shown that filter lists miss trackers. In this paper, we propose an alternative method to detect trackers inspired by analyzing behavior of invisible pixels. By crawling 84,658 webpages from 8,744 domains, we detect that third-party invisible pixels are widely deployed: they are present on more than 94.51% of domains and constitute 35.66% of all third-party images. We propose a fine-grained behavioral classification of tracking based on the analysis of invisible pixels. We use this classification to detect new categories of tracking and uncover new collaborations between domains on the full dataset of 4,216,454 third-party requests. We demonstrate that two popular methods to detect tracking, based on EasyList & EasyPrivacy and on Disconnect lists respectively miss 25.22% and 30.34% of the trackers that we detect. Moreover, we find that if we combine all three lists, 379,245 requests originated from 8,744 domains still track users on 68.70% of websites. This work will appear in PETS 2020 .

Participants: Mathieu Thiery, Arnaud Legout.  

The number of connected devices is increasing every day, creating smart homes and shaping the era of the Internet of Things (IoT), and most of the time, end-users are unaware of their impacts on privacy. In this work, we analyze the ecosystem around a Philips Hue smart white bulb in order to assess the privacy risks associated to the use of different devices (smart speaker or button) and smartphone applications to control it. We show that using different techniques to switch ON or OFF this bulb has significant consequences regarding the actors involved (who mechanically gather information on the user's home) and the volume of data sent to the Internet (we measured differences up to a factor 100, depending on the control technique we used). Even when the user is at home, these data flows often leave the user's country, creating a situation that is neither privacy friendly (and the user is most of the time ignorant of the situation), nor sovereign (the user depends on foreign actors), nor sustainable (the extra energetic consumption is far from negligible). We therefore advocate a complete change of approach, that favors local communications whenever sufficient. The preprint documenting this work has been published as research report .

Participants: Arnaud Legout, Mondi Ravi, David Migliacci, Abdelhakim Akodadi, Yanis Boussad.  

We are currently evaluating the relevance to create a startup for the ElectroSmart project. We are quite advanced in the process and the planned creation is June 2020. There is a "contrat de transfer" ready between Inria and ElectroSmart to transfer the PI from Inria to the ElectroSmart company (when it will be created). Arnaud Legout the future CEO of the company obtained the "autorisation de création d'entreprise" from Inria. ElectroSmart has been incubated in PACA Est in December 2018.

The three future co-founder of ElectroSmart (Arnaud Legout, Mondi Ravi, David Migliacci) followed the Digital Startup training from Inria/EM Lyon.

The goal of ElectroSmart is to help people reduce their exposure to EMF and offer a solution to reduce symptoms associated with exposure to EMF. Electrosensitivity, is known to be a complex and multifactorial syndrome that impacts hundreds of millions of persons worldwide. We aim to commercialize the first treatment of electrosensitivity based on non-deceptive placebo (called open-label placebo). It is known today that placebo are an effective treatment to subjective symptoms (which is the case for several symptoms associated with electrosensitivity). The problem with placebo was that is was assumed that it must be deceptive to be efficient. Kaptchuk et al. showed recently that non-deceptive placebo are as effective as deceptive placebo, so the ethical usage of placebo is now possible. ElectroSmart want to be the first company to commercialize non-deceptive placebo for electrosensitive persons. For details, see https://electrosmart.app/.

Participant: Damien Saucez.  

The objective of the ANR JCJC DET4ALL project was to offer the ability to multiplex constrained networks with real time and safety requirements on Ethernet network not initially thought for strict constraints. The reason for this move to Ethernet is to reduce the cost of networking solutions in automotive and industrial applications. We advocate that this move requires to rely on Software Defined Networking (SDN) that enables a programmatic approach to networking, hence offering modularity and flexibility. The challenge with SDN is to be able to certify the behaviour of the system while keeping the solution generic. Within DET4ALL we put the first element in place to show that the previous works that proposed programming languages and abstractions for best-effort network could be leveraged to offering safety properties and determinism in real-time industrial and automotive networks. More precisely, we have demonstrated that Linear Temporal Logic (LTL) can be used in real-time networks to demonstrate the that real-time constraints are always respected. We built a strawman to show that the Temporal NetKat language was adapted to express real-time constraints of networks even though it was not initially design for that purpose. Given that Temporal NetKat relies on LTL and an algebra, it is a good candidate to prove the correct behaviour of a SDN network which logic would be implemented with such a language. In the continuation of this work, we have determined what would be necessary to be able to provide provable live network updates in real time network without service degradation. This work is published in  and will be detailed in the next subsection. Due the leave of Damien Saucez to Safran for one year starting October 1st 2019, the activity on this project had to be stopped as it was in the context of an ANR JCJC project.

Participants: Damien Saucez, Walid Dabbous.  

Recent manufacturing trends have highlighted the need to adapt to volatile, fast-moving, and customer-driven markets. To keep pace with ever quicker product lifecycles, shorter order lead times and growing product variants, factories will become distributed modular cyber-physical systems interconnected by complex communication networks. We advocate that the Software Define Networking (SDN) concept with its programmatic approach to networking is a key enabler for the so-called Industry 4.0 because it provides flexibility and the possibility to formally reason on networks. We have identified that a critical point to address is how to support safe network updates of deterministic real-time communication SDN. To achieve this goal 4 elements are required. First a declarative language with LTL support is needed to express the constraints. Second, a programmable data-plane with the ability to provide real-time constraints indications must be provided in order to assess the behaviour of the forwarding elements. Such language does not exist yet however among the data-plane languages currently on the market some provide the ability to add annotations that could be used to reach our objective. Third, we have identified that deterministic algorithms had to be used to provide a verifiable sequence of network updates in order to make live updates without service degradations. Finally, mathematical techniques must be used to provide bounds on the network updates. Network Calculus can be used for that objective. This study was published as a poster in SOSR'19 .

Participants: Mohamed Naoufal Mahfoudi, Gayatri Sivadoss, Othmane Bensouda Korachi, Thierry Turletti, Walid Dabbous.  

We have proposed Snipe, a novel system offering joint localization and range extensions for LPWANs. Although LPWAN systems such as Long Range (LoRa) are designed to achieve high communication range with low energy consumption, they suffer from fading in obstructed environments with dense multipath components, and their localization system is sub-par in terms of accuracy. In this work, MIMO techniques are leveraged to achieve a higher signal-to-noise ratio at both the end device and the gateway while providing an opportunistic accurate radar-based system for localization with limited additional cost. This work has been published at Internet Technology Letters .

Participants: Ghada Moualla, Thierry Turletti, Damien Saucez.  

The trend today is to deploy applications and more generally Service Function Chains (SFCs) in public clouds. However, before being deployed in the cloud, chains were deployed on dedicated infrastructures where software, hardware, and network components were managed by the same entity, making it straightforward to provide robustness guarantees. By moving their services to the cloud, the users lose their control on the infrastructure and hence on the robustness. We propose an online algorithm for robust placement of service chains in data centers. Our placement algorithm determines the required number of replicas for each function of the chain and their placement in the data center. Our simulations on large data-center topologies with up to 30,528 nodes show that our algorithm is fast enough such that one can consider robust chain placements in real time even in a very large data center and without the need of prior knowledge on the demand distribution. This work has been published at IEEE Access .

Participants: Giuseppe Di Lena, Damien Saucez, Thierry Turletti.  

With the emergence of Network Function Virtualization (NFV) and Software Defined Networking (SDN), efficient network algorithms considered too hard to be put in practice in the past now have a second chance to be considered again. In this context, we rethink the network dimensioning problem with protection against Shared Risk Link Group (SLRG) failures. In this work, we consider a path-based protection scheme with a global rerouting strategy, in which, for each failure situation, there may be a new routing of all the demands. Our optimization task is to minimize the needed amount of bandwidth. After discussing the hardness of the problem, we develop a scalable mathematical model that we handle using the Column Generation technique. Through extensive simulations on real-world IP network topologies and on random generated instances, we show the effectiveness of our method. Finally, our implementation in OpenDaylight demonstrates the feasibility of the approach and its evaluation with Mininet shows that technical implementation choices may have a dramatic impact on the time needed to reestablish the flows after a failure takes place. This work has been presented at the IEEE International Conference on Cloud Networking (CloudNet), November 2019, at Coimbra in Portugal  and documented in a research report . A poster version is published in IFIP-Networking in Warsaw .

Participants: Indukala Naladala, Thierry Turletti.  

There has been a considerable increase in the demand for high quality mobile video streaming services, while at the same time, the video traffic volume is expected to grow exponentially. Consequently, maintaining high quality of experience (QoE) and saving network resources are becoming crucial challenges to solve. In this work, we propose a name-based mobile streaming scheme that allows efficient video content delivery by exploiting a smart pulling mechanism designed for information-centric networks (ICNs). The proposed mechanism enables fast packet loss recovery by leveraging in-network caching and coding. Through an experimental evaluation of our mechanism over an open wireless testbed and the Internet, we demonstrate that the proposed scheme leads to higher QoE levels than classical ICN and TCP-based streaming mechanisms. This work will be presented at the IEEE Consumer Communications & Networking Conference (CCNC), in January 2020 at Las Vegas, USA . The following link https://github.com/fit-r2lab/demo-cefore includes the artefacts that allows to reproduce performance results shown in the paper.

Participants: Luigi Vigneri, Chadi Barakat.

Caching content at the edge of mobile networks is considered as a promising way to deal with the data tsunami. In addition to caching at fixed base stations or user devices, it has been recently proposed that an architecture with public or private transportation acting as mobile relays and caches might be a promising middle ground. While such mobile caches have mostly been considered in the context of delay tolerant networks, in this work done in collaboration with Eurecom with the support of the UCN@Sophia Labex, we argue that they could be used for low cost video streaming without the need to impose any delay on the user. Users can prefetch video chunks into their playout buffer from encountered vehicle caches (at low cost) or stream from the cellular infrastructure (at higher cost) when their playout buffer empties while watching the content. Our main contributions are: (i) to model the playout buffer in the user device and analyze its idle periods which correspond to bytes downloaded from the infrastructure; (ii) to optimize the content allocation to mobile caches, to minimize the expected number of non-offloaded bytes. We perform trace-based simulations to support our findings showing that up to 60 percent of the original traffic could be offloaded from the main infrastructure. These contributions were published in the IEEE Transactions on Mobile Computing journal .

Participants: Luigi Vigneri, Chadi Barakat.  

Densification through small cells and caching in base stations have been proposed to deal with the increasing demand for Internet content and the related overload on the cellular infrastructure. However, these solutions are expensive to install and maintain. Instead, using vehicles acting as mobile caches might represent an interesting alternative. In this work, we assume that users can query nearby vehicles for some time, and be redirected to the cellular infrastructure when the deadline expires. Beyond reducing costs, in such an architecture, through vehicle mobility, a user sees a much larger variety of locally accessible content within only few minutes. Unlike most of the related works on delay tolerant access, we consider the impact on the user experience by assigning different retrieval deadlines per content. We provide the following contributions: (i) we model analytically such a scenario; (ii) we formulate an optimization problem to maximize the traffic offloaded while ensuring user experience guarantees; (iii) we propose two variable deadline policies; (iv) we perform realistic trace-based simulations, and we show that, even with low technology penetration rate, more than 60% of the total traffic can be offloaded which is around 20% larger compared to existing allocation policies. These results were published in the IEEE Transactions on Mobile Computing journal .

Participants: Tingting Yuan, Thierry Turletti, Chadi Barakat.

Intelligent Transportation Systems, or ITS for short, includes a variety of services and applications such as road traffic management, traveler information systems, public transit system management , and autonomous vehicles, to name a few. It is expected that ITS will be an integral part of urban planning and future cities as it will contribute to improved road and traffic safety, transportation and transit efficiency, as well as to increased energy efficiency and reduced environmental pollution. On the other hand, ITS poses a variety of challenges due to its scalability and diverse quality-of-service needs, as well as the massive amounts of data it will generate. In this survey, we explore the use of Machine Learning (ML), which has recently gained significant traction, to enable ITS. In the context of the Drive associated team, we did a comprehensive survey of the current state-of-the-art of how ML technology has been applied to a broad range of ITS applications and services, such as cooperative driving and road hazard warning, and identify future directions for how ITS can use and benefit from ML technology. The survey is documented in .

Participants: Giuseppe Di Lena, Damien Saucez, Thierry Turletti.  

Networks have become complex systems that combine various concepts, techniques, and technologies. As a consequence , modelling or simulating them is now extremely complicated and researchers massively resort to prototyping techniques. Among other tools, Mininet is the most popular when it comes to evaluate SDN propositions. It allows to emulate SDN networks on a single computer. However, under certain circumstances experiments (e.g., resource intensive ones) may overload the host running Mininet. To tackle this issue, we propose Distrinet, a way to distribute Mininet over multiple hosts. Distrinet uses the same API than Mininet, meaning that it is compatible with Mininet programs. Distrinet is generic and can deploy experiments in Linux clusters or in the Amazon EC2 cloud. Thanks to optimization techniques, Distrinet minimizes the number of hosts required to perform an experiment given the capabilities of the hosting infrastructure, meaning that the experiment is run in a single host (as Mininet) if possible. Otherwise, it is automatically deployed on a platform using a minimum amount of resources in a Linux cluster or with a minimum cost in Amazon EC2. This work has been presented at the IEEE International Conference on Cloud Networking (CloudNet) . Distrinet has been demonstrated both at the IEEE CloudNet conference and at the ACM CoNEXT conference in Orlando USA in December 2019 .

Participants: Giuseppe Di Lena, Damien Saucez, Thierry Turletti, Walid Dabbous.  

With the ever growing complexity of networks, researchers have to rely on test-beds to be able to fully assess the quality of their propositions. In the meanwhile, Mininet offers a simple yet powerful API, the goldilocks of network emulators. We advocate that the Mininet API is the right level of abstraction for network experiments. Unfortunately it is designed to be run on a single machine. To address this issue we developed a distributed version of Mininet-Distrinet-that can be used to perform network experiments in any Linux-based testbeds, either public or private. To properly use testbed resources and avoid over-commitment that would lead to inaccurate results, Distrinet uses optimization techniques that determine how to orchestrate the experiments within the testbed. Its programmatic approach, its ability to work on various testbeds, and its optimal management of resources make Distrinet a key element to reproducible research. This work has been presented at the Global Experimentation for Future Internet - Workshop (GeFi) workshop November 2019, at Coimbra in Portugal .

Participants: Yanis Boussad, Arnaud Legout.  

From crowdsource data collection to automation and robotics, mobile smartphones are well suited for various use cases given the rich hardware components they feature. Researchers can now have access to various sensors such as barometers, magnetometers, orientation sensors, in addition to multiple wireless technologies all on a single and relatively cheap mobile smartphone. In this work, we study the performance of smartphones to measure cellular wireless power. We performed our experiments inside an anechoic chamber in order to compare the measurements of smartphone to the ones obtained with professional spectrum analyzer. We first evaluate the effect of orientation on the received power, then we propose a way to improve the accuracy of smartphone power measurements by using the orientation sensors. We improve the accuracy of the measurements from 25 dBm RMSE to no more than 6 dBm RMSE. We also show how we can exploit the characteristics of the reception pattern of the smartphone to determine the angle of arrival of the signal. The results of this work are described in a research report under submission .

Participants: Mohamed Naoufal Mahfoudi, Thierry Parmentelat, Thierry Turletti, Walid Dabbous.  

Reproducibility is key in designing wireless systems and evaluating their performance. Trying to reproduce wireless experiments allowed us to identify some pitfalls and possible ways to simplify the complex task of avoiding them. In this research report, we expose a few considerations that we learned are instrumental for ensuring the reproducibility of wireless experiments. Then we describe the steps we have taken to make our experiments easy to reproduce. We specifically address issues related to wireless hardware, as well as varying propagation channel conditions. We show that extensive knowledge of the used hardware and of its design is required to guarantee that the inner state of the system has no negative impact on performance evaluation and experimental results. As for variability of channel conditions, we make the case that a special setup or testbed is necessary so that one can control the ambient wireless propagation environment, using for instance, an anechoic chamber like R2lab. This work is published as research report .

Participants: Thierry Parmentelat, Mohamed Naoufal Mahfoudi, Thierry Turletti, Walid Dabbous.  

In this research report, we present R2lab, an open, electromagnetically insulated research testbed dedicated to wireless networking. We describe the hardware capabilities currently available in terms of Software Defined Radio, and the software suite made available to deploy experiments. Using a generic experiment example, we show how it all fits into a notebook-based approach to getting closer to runnable papers. This work is published as research report .