The ability to search databases by providing multimedia examples of voices, faces, or locations instead of textual descriptions can be tremendously useful. At the same time, uploading media for queries---especially media that contains sensitive content---means sharing private information with a potentially untrusted service provider. The growing field of privacy-preserving database searches attempts to resolve this tension. Within this scope of private searches, private media classification and retrieval is particularly challenging due to the inherent inexactness of recognition; to be useful, image or other media classification systems must identify approximate matches rather than just exact ones. This is difficult to reconcile with distortion-intolerant and resource-heavy privacy primitives, especially in web-scale databases. In this paper, we present an architecture for media classification on public databases that preserves client privacy while achieving asymptotic communication and computation costs that are sublinear in the size of the database. We demonstrate the usefulness of this architecture in the context of a privacy-preserving face recognition system. We observe order-of-magnitude speedups over state-of-the-art private face recognition systems.

This article presents a new algorithm to find MDS matrices that are well suited for use as a diffusion layer in lightweight block ciphers. Using an recursive construction, it is possible to obtain matrices with a very compact description. Classical field multiplications can also be replaced by simple F2-linear transformations (combinations of XORs and shifts) which are much lighter. Using this algorithm, it was possible to design a 16×16 matrix on a 5-bit alphabet, yielding an efficient 80-bit diffusion layer with maximal branch number.

Private Stream Search allows users to perform keyword-based queries to a database without revealing any information about the keywords they are searching. Using homomorphic encryption, Ostrovsky and Skeith proposed a computationally secure solution to this problem in 2005. However, their solution requires the server to send an answer of size O(mS log m) bits when m documents of S bits match the query, while a non-private query only requires mS bits. In this work we propose two new communication optimal constructions, both allowing a communication expansion factor (compared to a non-private query) asymptotically equal to 1 when m and S increase. More precisely, our first scheme requires m(S + O(log t)) bits (where t is the size of the database) and our second scheme m(S +C) where C is a constant depending on the chosen computational security level.

We present two new algorithms for the reconstruction of turbo codes from a noisy intercepted bitstream. With these algorithms, we were able to reconstruct various turbo codes with realistic parameter sizes. To the best of our knowledge, these are the first algorithms able to recover the whole permutation of a turbo code in the presence of high noise levels.

We focus on the problem of recovering the length and synchronization of a linear block code from an intercepted bitstream. We place ourselves in an operational context where the intercepted bitstream contains a realistic noise level. We present two algorithms, one due to Valembois and the other one brand new. They are both useful in different contexts, able to verify if a given length/synchronization is correct. Using them, we were able to practically recover the synchronization of several codes.

We propose a method to model and control topological changes by a smooth deformation of a polyhedral mesh using curves and loops. As changing the genus of a surface is not a continuous transformation, the topological change is made when an intermediate shape between the two topologies has been obtained. The creation and the deletion of holes are studied. The deletion of a hole uses non null-homotopic loops to designate the hole to be deleted. A method computing two independent loops associated to a hole is presented.