Section: New Results

Privacy

The right to be forgotten, or to oblivion, is an aspect of privacy rights. It relates to the need for individuals to be able to leave a part of their past behind them, to change their mind about something or to take a new start in a given domain. The final report of the DAO project [53] presents an analysis of the concept from a multidisciplinary point of view, including a sociological study, a legal state of the art assorted with insights of possible evolutions, and a technical state of the art along with the proposal of a new architecture [22] . A joint technical and legal analysis of the conceptual and technical issues specific to social networks is also proposed. From the point of view of a computer scientist, the most obvious issue with the right to be forgotten is the ability to control the deletion of a piece of information once it has been disclosed and disseminated. In the general case, no guarantees can be provided, but under certain conditions it is possible to enforce remote deletion with reasonable guarantees. In general, it implies that architectural and applicative choices are made beforehand, either to allow for future decisions regarding data made available in a controlled framework, like late modifications of its access policy or the triggering of its destruction, or to plan deletion from the beginning and set a time-to-leave when disclosing the data within a particular environment, or . The approach designed in CIDRE, relying on both ephemeral publication and data degradation techniques, falls in the latter category, improving the utility for third parties (when compared to existing ephemeral publication techniques) and building a new trade-off with the users' privacy needs, by making different versions of the original data, more or less precise, available for different durations, the more detailed information being lost the quickliest.

CIDRE also contributes, through the B<>com IRT, to the supervision (by Annie Blandin, professor at Télécom Bretagne, and Guillaume Piolle) of Gustav Malis's doctoral work in law in the domain of a restrictive case of the right to be forgotten. In this context, very original contributions have been made at the intersection between the two fields. In particular, a joint analysis has been proposed on the roles of legal and computing tools for the implementation of the right to be forgotten [50] . In particular, it seems that the two domains consider the issue with very different perspectives: the computer scientist almost takes for granted that he cannot rely on regulations and on “security through legality”, hence the tools he designs are intended to directly empower the user, putting him in control of his data by using preventive protection techniques. The tools may fail though, or more likely their applicability conditions may not suit all scenarios. When issues arise they may be captured by the regulatory framework, which intends to provide means for reparation and restoration. Both approaches fail to encompass all possible situations and to solve all potential issues, but they provide users and citizens with complementary tools.

The work combining computer science and law conducted in the DAO projet as well as the main conclusions of the project have also been presented in interdisciplinary colloquium by Sébastien Gambs and Maryline Boizard [48] , [47] .

Mobility has always been an important aspect of human activities. Nowadays problems of congestion in urban areas due to the massive usage of cars, last-minutes travel needs and progress in information and communication technologies encourage the rise of new transport modes. Among those are carpooling services, which let car owners share the empty seats of their cars with other travellers having the same travel destination. However, the way carpooling services are implemented today raises several privacy issues. In a recent paper, together with researchers from LAAS-CNRS we have proposed to use privacy enhancing technologies to improve the quality of carpooling services by specially taking in consideration privacy aspects [46] .

In addition, publishing directly human mobility data raises serious privacy issues due to its inference potential, such as the (re-)identification of individuals. To address these issues and to foster the development of such applications in a privacy-preserving manner, we propose in a recent paper [26] a novel approach in which Call Detail Records (CDRs) are summarized under the form of a differentially-private Bloom filter for the purpose of privately estimating the number of mobile service users moving from one area (region) to another in a given time frame. Our sanitization method is both time and space efficient, and ensures differential privacy while solving the shortcomings of a solution recently proposed. We also report on experiments conducted using a real life CDRs dataset, which show that our method maintains a high utility while providing strong privacy.

Finally, in authentication protocols, a relay attack allows an adversary to impersonate a legitimate prover, possibly located far away from a verifier, by simply forwarding messages between these two entities. The effectiveness of such attacks has been demonstrated in practice in many environments, such as ISO 14443-compliant smartcards and car-locking mechanisms. Distance-bounding (DB) protocols, which enable the verifier to check his proximity to the prover, are a promising countermeasure against relay attacks. In such protocols, the verifier measures the time elapsed between sending a challenge and receiving the associated response of the prover to estimate their proximity. So far, distance bounding has remained mainly a theoretical concept. Indeed in practice, up to our knowledge only three ISO 14443-compliant implementations of DB protocols exist. The first two are implemented on proprietary smartcards while the last one is available on a highly-customized and dedicated hardware. In a recent paper [35] , we demonstrated a proof-of-concept implementation of the Swiss-Knife DB protocol on smartphones running in RFID-emulation mode. To our best knowledge, this is the first time that such an implementation has been performed. Our experimental results are encouraging as they show that relay attacks introducing more than $1.5$ ms are directly detectable (in general off-the-shelf relay attacks introduce at least 10 ms of delay). We also leverage on the full power of the ISO-DEP specification to implement the same protocol with 8-bit challenges and responses, thus reaching a better security level per execution without increasing the possibility of relay attacks. The analysis of our results leads to new promising research directions in the area of distance bounding.