Section: New Results

Highlights of the Year

The supervision of distributed system relies heavily on correlation mechanisms that are responsible for collecting alerts coming from sensors and detecting complex scenarios in the flow of alerts. The problem is that it requires to write complex correlation rules. The work we have performed proposes a technique to generate semi-automatically such correlation rules. It describes a process that uses an attack tree and a representation of the system as inputs, and generate a correlation tree that can be translated in an alert correlation description language. This work received the best paper award of SAR-SSI 2014 [50] .

One approach to protect the privacy of users in personalized recommendation systems is to publish a sanitized version of the profile of the user by relying a non-interactive mechanism compliant with the concept of differential privacy. In a joint work with Raghavendran Balu and Teddy Furon (LinkMedia Inria team), we have consider two existing schemes offering a differentially private representation of profiles: BLIP (BLoom-and-flIP) and JLT (Johnson-Lindenstrauss Transform). For assessing their security levels, we play the role of an adversary aiming at reconstructing a user profile. To realize this, we design two inference attacks named single and joint decoding. The first inference attack tests the presence of a single item in the profile, and is iterated independently for each possible item of the item set. In contrast, the second inference attack aims at deciding whether a particular subset of items is likely to be in the user profile. This attack is tested on all the possible subsets of items. Our contributions are a theoretical analysis and practical implementations of both attacks tested on datasets composed of real user profiles revealing that joint decoding is the most powerful attack. This also gives useful insights on the setting the differential privacy parameter $ϵ$. This work has received the best student paper award at the conference ESORICS 2014.

Best Paper Award :