La prochaine édition du séminaire C2 se tiendra le vendredi 15 novembre 2019 en salle Jaques-Louis Lions 2 de l’INRIA Paris, 2 ,rue Simone Iff, Métro Ligne 6, arrêt Dugommier.
Exceptionnellement, le séminaire ne se déroulera que le matin, la thèse de Xavier Bonnetain se déroulant l’après-midi également à 14h00 : Xavier Bonnetain, Inria Paris, Soutenance de thèse : Hidden Structures and Quantum Cryptanalysis
- 10h00 -Geoffroy Couteau, IRIF, Université Paris-Diderot : Efficient Pseudorandom Correlation Generators: Silent OT Extension and More
Abstract : I will present recent developments in secure computation in the preprocessing model. In this model, a protocol is divided in two phases: a preprocessing phase, independent of the inputs, during which long correlated strings are generated and distributed among the parties; and a lightweight online phase, were the actual computation takes place. The generation of the preprocessing material forms the main efficiency bottleneck of all modern secure computation protocols, due to the very large communication, computation, and storage involved.
In this talk, I will discuss how a new cryptographic primitive, a pseudorandom correlation generator (PCG), can be used to considerably improve this state of affair. Informally, a PCG allows to stretch short, correlated keys into long correlated pseudorandom string of near-arbitrary length. This allows the preprocessing phase of all secure computation protocols to be executed with a tiny amound of communication (to generate the short correlated keys) followed only by local operations (to stretch correlated strings from these keys). I will discuss how to formalize this primitive, and describe a way to construct PCGs for an important class of correlations from the hardness of syndrome decoding. In turns, this results in considerable improvements for state-of-the-art secure computation, both from an asymptotic point of view and in practice.
- 11h15 – Pierre Karpman, Université Grenoble Alpes : High-order private multiplication in characteristic two revisited
Abstract : We revisit the high-order masking schemes for private multiplication introduced by Belaïd et al. at EUROCRYPT 2016, and the matrix model for non-interference (NI) security that they develop in their follow-up work of CRYPTO 2017. This leads to two main results.
1) We generalise the theorems of CRYPTO 2017 so as to be able to apply them to masking schemes over any finite field — in particular GF(2) — and to be able to analyse the strong non-interference (SNI) security notion. This leads to an efficient algorithm that allows us to computationally check the (S)NI security of concrete multiplication gadgets over GF(2) up to order 11.
2) We propose new SNI and NI multiplication gadgets over GF(2) (and any extension thereof) up to order 9 and 11 that moderately improve the randomness complexity of the schemes of Barthe et al. (EUROCRYPT 2017).
