Vous êtes ici : Accueil » Kiosque » Annonce

Identification

Identifiant: 
Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?

Annonce

16 février 2021

Deep Reinforcement Learning-based location-privacy preserving for large networks


Catégorie : Stagiaire


 

Context:

Location privacy has been becoming more and more important in wireless networks due to a large amount of data sharing, especially with the emergence of location sharing applications, and then the concept of location-based services is derived. Indeed, in such services, the users constantly report back to the infrastructure controller sensitive information such as their location. Nevertheless, leaking the location to a third-party introduces a privacy issue. Therefore, how to preserve location privacy is a key issue in location-based services. Indeed, two categories of location privacy-preserving schemes are presented in [1]: privacy enhancing schemes, like mix zones approaches that enhance privacy using anonymous communication zones, and privacy preserving authentication schemes that are based on symmetric or asymmetric key cryptography.Indeed, there are different ways to protect location privacy of users; one solution consists in allowing users to disseminate obfuscated locations, to trick potential malicious parties, which is the least stringent in terms of trust requirement among users, and the least computationally demanding. In [2], [3], obfuscation schemes based on a randomization of locations via sporadic perturbation are proposed. In particular, the proposed scheme in [3], adds noise drawn from a Laplace distribution such that the differential privacy is guaranteed. Unfortunately, such an assumption is not very relevant because the performance using the new position is not evaluated, and therefore the perturbation could not be adapted in order to guarantee a certain quality of service. Moreover, the major drawback of these methods is that they cannot scale to a large network (e.g. heavy traffic) without unaffordable complexity increase. The goal of this internship is to overcome these limitations and propose adaptive scheme for location-privacy in large networks.

Objectives:

The main objectives of this internship are:

1.Conduct a survey about state-of-the art methods for location-privacy preservation.

2.Using reinforcement learning provides encouraging results [4], in terms of privacy-preserving, while keeping the quality of service at a satisfying level.However, one major limitation to this solution is that the complexity grows drastically with the problem size. Therefore, the problem cannot be modeled efficiently anymore as a classical reinforcement learning framework.The second objective of this internship is to develop a deep reinforcement learning framework to model efficiently and learn the high-dimensional involved parameters in location based services within wireless networks.

3.In addition, tackle a more complex scenario, of a large size network by investigating and optimizing advanced machine learning architectures.

Skills:

1.The candidate should be enrolled in a master (or bac + 5) in wireless communications, electrical engineering or any related field.

2.Good knowledge of Matlab and/or Python.

3.Knowledge of machine learning is a plus.

Location:

1.Adresse : ENSEA, 6 Rue du Ponceau, 95000 Cergy.

2.Research team : ICI group, ETIS UMR 8051 (CY Cergy Paris University, ENSEA, CNRS)

3.Duration: 5-6 months.

 

Application:

CV and cover letter should be sent to Sara Berri (sara.berri@ensea.fr ) and Marwa Chafii (marwa.chafii@ensea.fr ).

 

References:

[1] P. Asuquo, H. Cruickshank, J. Morley, C. P. A. Ogah, A. Lei, W. Hathal, S. Bao, and Z. Sun, Security and Privacy in Location-Based Services for Vehicular and Mobile Communications: An Overview, Challenges, and Countermeasures, IEEE Internet Of Things Journal, vol. 5, no. 6, pp. 4778-4802, Dec. 2018.

[2] K. Emara, W. Woerndl, J. Schlichter, On Evaluation Of Location Privacy Preserving Schemes For VANET Safety Applications, Computer Communications, vol. 15, pp. 11-23, 2015.

[3] J. Zhang, Q. Yang, Y. Shen, Y. Wan, A Differential Privacy Based Probabilistic Mechanism for Mobility Datasets Releasing, Journal of Ambient Intelligence and Humanized Computing, pp. 1-12, 2020.

[4] S. Berri, J. Zhang, B. Bensaou, and H. Labiod, Privacy-Preserving Data-Prefetching in Vehicular Networks via Reinforcement Learning, In Proc. IEEE International Conference on Communications (ICC), Virtual Conference, 07-11 June. 2020.

 

Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2020.