Section: New Results

Personalization and Privacy

In work [37] we propose fingerprinting, a new technique that consists in constructing compact, fast-to-compute and privacy-preserving representation of datasets. We illustrate the effectiveness of our approach on the emblematic big data problem of K-Nearest-Neighbor (KNN) graph construction and show that fingerprinting can drastically accelerate a large range of existing KNN algorithms, while efficiently obfuscating the original data, with little to no overhead. Our extensive evaluation of the resulting approach (dubbed GoldFinger) on several realistic datasets shows that our approach delivers speed-ups up to 78.9% compared to the use of raw data while only incurring a negligible to moderate loss in terms of KNN quality. To convey the practical value of such a scheme, we apply it to item recommendation, and show that the loss in recommendation quality is negligible.

This work was done in collaboration with Rachid Guerraoui (EPFL) and Anne-Marie Kermarrec (Mediego/EPFL).

Recommendation systems help users identify interesting content, but they also open new privacy threats. For this reason, in [22] we deeply analyzed the effect of a Sybil attack that tries to infer information on users from a user-based collaborative-filtering recommendation systems. We evaluated the impact of different similarity metrics used to identity users with similar tastes in the trade-off between recommendation quality and privacy. Based on our results, we proposed and evaluated a novel similarity metric that combines the best of both worlds: a high recommendation quality with a low prediction accuracy for the attacker. Our experiments, on a state-of-the-art recommendation framework and on real datasets showed that existing similarity metrics exhibit a wide range of behaviors in the presence of Sybil attacks, while our new similarity metric consistently achieves the best trade-off while outperforming state-of-the-art solutions.

This work was carried out in collaboration with Antoine Boutet from INSA Lyon, former-intern Florestan De Moor, Rachid Guerraoui and Antoine Rault from EPFL, and Anne-Marie Kermarrec from Mediego.