(French version below)

Note: A video recording of the defense as well as the slides of the presentation can be found here.

You are cordially invited to attend the thesis defense of Antoine Rault that will be held on Thursday June 23, 2016 at 14:30 in room Métivier and for a drink in room Minquiers.

Title: User Privacy in Collaborative Filtering Systems

Jury members:
Ludovic Mé, Professor, Supelec
Pascal Felber, Professor, Université de Neuchâtel
Arnaud Legout, Researcher, Inria Sophia Antipolis
Patrick Loiseau, Assistant professor, Eurocom
Sébastien Gambs, Professor, Université du Québec, Montréal
Davide Frey, Researcher, Inria Rennes Bretagne Atlantique
Anne-Marie Kermarrec, Thesis supervisor, Mediego

Abstract:
Recommendation systems try to infer their users’ interests in order to suggest items relevant to them. These systems thus offer a valuable service to users in that they automatically filter non-relevant information, which avoids the nowadays common issue of information overload. This is why recommendation systems are now popular, if not pervasive in some domains such as the World Wide Web. However, an individual’s interests are personal and private data, such as one’s political or religious orientation. Therefore, recommendation systems gather private data and their widespread use calls for privacy-preserving mechanisms. In this thesis, we study the privacy of users’ interests in the family of recommendation systems called Collaborative Filtering (CF) ones. Our first contribution is Hide & Share, a novel privacy-preserving similarity mechanism for the decentralized computation of K-Nearest-Neighbor (KNN) graphs. It is a lightweight mechanism designed for decentralized (a.k.a. peer-to-peer) user-based CF systems, which rely on KNN graphs to provide recommendations.
Our second contribution also applies to user-based CF systems, though it is independent of their architecture. This contribution is two-fold: first we evaluate the impact of an active Sybil attack on the privacy of a target user’s profile of interests, and second we propose a counter-measure. This counter-measure is 2-step, a novel similarity metric combining a good precision, in turn allowing for good recommendations, with high resilience to said Sybil attack.


Note: Un enregistrement vidéo de la soutenance ainsi que le support de la présentation sont disponibles ici.

Vous êtes cordialement invités à venir assister à la soutenance de thèse de Antoine Rault qui se tiendra jeudi 23 juin 2016 à 14h30 en salle Métivier ainsi qu’au pot qui suivra en salle Minquiers.

Titre : Protection de la vie privée des utilisateurs de systèmes de filtrage collaboratif

Composition du jury :
Ludovic Mé, Professeur, Supelec Rennes
Pascal Felber, Professeur, Université de Neuchâtel
Arnaud Legout, Chargé de recherche, Inria Sophia Antipolis
Patrick Loiseau, Maître de Conférence, Eurocom, Sophia Antipolis
Sébastien Gambs, Professeur, Université du Québec, Montréal
Davide Frey, Chargé de recherche, Inria Rennes Bretagne Atlantique
Anne-Marie Kermarrec, Directrice de thèse, Mediego

Résumé :
Les systèmes de recommandation essayent de déduire les intérêts de leurs utilisateurs afin de leurs suggérer des items pertinents. Ces systèmes offrent ainsi aux utilisateurs un service utile car ils filtrent automatiquement les informations non-pertinentes, ce qui évite le problème de surcharge d’information qui est courant de nos jours. C’est pourquoi les systèmes de recommandation sont aujourd’hui populaires, si ce n’est omniprésents dans certains domaines tels que le World Wide Web. Cependant, les intérêts d’un individu sont des données personnelles et privées, comme par exemple son orientation politique ou religieuse. Les systèmes de recommandation recueillent donc des données privées et leur utilisation répandue nécessite des mécanismes de protection de la vie privée. Dans cette thèse, nous étudions la protection de la confidentialité des intérêts des utilisateurs des systèmes de recommandation appelés systèmes de filtrage collaboratif (FC). Notre première contribution est Hide & Share, un nouveau mécanisme de similarité, respectueux de la vie privée, pour le calcul décentralisé de graphes de K-Plus-Proches-Voisins (KPPV). C’est un mécanisme léger, conçu pour les systèmes de FC fondés sur les utilisateurs et décentralisés (ou pair-à-pair), qui se basent sur les graphes de KPPV pour fournir des recommandations.
Notre seconde contribution s’applique aussi aux systèmes de FC fondés sur les utilisateurs, mais est indépendante de leur architecture. Cette contribution est double : nous évaluons d’abord l’impact d’une attaque active dite « Sybil » sur la confidentialité du profil d’intérêts d’un utilisateur cible, puis nous proposons une contre-mesure. Celle-ci est 2-step, une nouvelle mesure de similarité qui combine une bonne précision, permettant ensuite de faire de bonnes recommandations, avec une bonne résistance à l’attaque Sybil en question.

We are receiving Daniel Negru and Joachim Bruneau-Queyreix from the LaBRI, University of Bordeaux, this week (9 and 10 of June).

In this context, Thursday, June 9 at 14:00 in room Minquiers (B025), we will have two talks: one by Joachim Bruneau-Queyreix and one by Davide Frey.

  • Joachim Bruneau-Queyreix
  • Title: DASH evolution for Quality of Experience Enhancement via On-Demand Multiple-Server Support: A game-changer for future content delivery solution

    Abstract:
    Adaptive bitrate streaming protocols, such as DASH, have seen extensive interest for their adaptation capabilities towards end-users Quality of Experience (QoE) increase. In this talk, we will present our solution to enrich DASH potential with multiple-server features and to take advantage of expanded bandwidth, link diversity, and reliability. Thanks to its codec agnosticism, DASH-compliance, and receiver-driven philosophy, our contribution is a pragmatic and evolving solution for QoE enhancement that can be applied to many streaming architectures (P2P, CDNs, Clouds). By splitting content into multiple independent and aggregatable sub-streams, we achieve easy-to-design content- and server- adaptation mechanisms for high QoE increase. We empirically validate our approach using an extensive collection of network profiles provided by the DASH Industry Forum. Comparison with the full potential of DASH in a multi-source environment is also performed over several criteria, resulting in very important QoE gains.
    Enabling multiple-server adaptive streaming solutions over HTTP represents a key feature in bringing new content delivery solutions (i.e. Set-Top Box overlays, collaboration of multiple actors of the content delivery chain, CDN/Cloud/peer-based solutions) to the market for high and ultra-high definition streaming (VoD, OTT-TV, UGC, UGC-broadcasting events). Further discussion on potential impacts of multiple-server DASH in existing and future content delivery solutions will be held. An online demonstration of this work will also be presented.

  • Davide Frey
  • Title: Live Streaming with Gossip

    Abstract:
    Video streaming has become the new killer application for peer-to-peer technologies. By aggregating scarce resources such as upload bandwidth, decentralized video streaming protocols make it possible to serve a video stream to huge numbers of users while requiring very limited investments from broadcasters. In this talk, we discuss gossip-based video streaming, and present HEAP, Heterogeneity-Aware gossip Protocol. HEAP (HEterogeneity-Aware gossip Protocol) incorporates several features that improve the standard gossip-based protocols used in video-streaming applications. First it includes a fanout adaptation mechanism that tunes the contribution of nodes to the streaming process based on their bandwidth capability. Second, it comprises heuristics that improve reliability, as well as operation in the presence of heterogeneous network latency. We evaluate HEAP on a real deployment on the Grid5000 platform. Results show that HEAP significantly improves the quality of the streaming over standard homogeneous gossip protocols, especially when the stream rate is close to the average available bandwidth.

Our paper “The Out-of-core KNN awakens: The light side of computation force on large datasets” by Nitin Chiluka, Anne-Marie Kermarrec and Javier Olivares has received the Best Paper Award at The 4th Edition of The International Conference on NETworked sYStems (NETYS 2016).

We are receiving Antonio Carzaniga from USI Switzerland (Università della Svizzera Italiana). He will give a talk on Friday, April 29 at 14:00 in room Aurigny.

Title: Descriptors, Locators, Identifiers: Multi-Modal Addressing in the TagNet Information-Centric Networking Architecture

Abstract:
A truly information centric network is one where addresses are given by applications (users), not by the network. Application-defined addresses make applications easier to write and to deploy, but on the other and such addresses might not aggregate very well, which then limits the scalability of routing and forwarding. This is a crucial dilemma: how to architect a network so that applications can benefit from expressive and flexible addressing, and at the same time the network can scale. In this talk I will present the approach we took to solve this dilemma within the TagNet architecture. TagNet supports two distinct types of addresses at the network level: descriptors and locators. Descriptors are expressive and are fully application-defined, while locators are opaque network-level addresses. TagNet also distinguishes the role of identifiers, which play a fundamental role in caching and transport protocols, but not in routing and forwarding. With these three concepts, TagNet supports both push and pull communication, and, I will argue, it can do that with scalable routing and forwarding.

Bio:
Antonio Carzaniga is a professor in the Faculty of Informatics at USI, Switzerland (Università della Svizzera italiana), which he joined as an assistant professor when the Faculty was founded in 2004. From 2001 to 2007 he was also an assistant research professor in the Department of Computer Science at the University of Colorado at Boulder. Antonio Carzaniga received the Laurea degree in electronics engineering and the Ph.D. degree in computer science from Politecnico di Milano, Italy. Antonio Carzaniga’s primary research interests are in the fields of distributed systems and software engineering, and more specifically in content-based networking, information-centric networking, distributed publish/subscribe systems, middleware, software adaptability and automatic fault tolerance, and testing. He also conducted research in software configuration management and code mobility.

Antonio’s personal page: http://www.inf.usi.ch/carzaniga

Our paper “The Out-of-core KNN awakens: The light side of computation force on large datasets” by Nitin Chiluka, Anne-Marie Kermarrec and Javier Olivares has been accepted at The 4th Edition of The International Conference on NETworked sYStems (NETYS 2016).

« Newer Posts - Older Posts »