Sommaire

  • Cet exposé a été présenté le 13 mars 2026 (13:45 - 14:45).

Description

  • Orateur

    Yevhen Perehuda - Ruhr-University Bochum

In this work, we address the critical yet understudied question of the security of the most widely deployed pseudorandom number generators (PRNGs) in AI applications. We show that these generators are vulnerable to practical and low-cost attacks. With this in mind, we conduct an extensive survey of randomness usage in current applications to understand the efficiency requirements imposed in practice. Finally, we present a cryptographically secure and well-understood alternative, which has a negligible effect on the overall AI/ML workloads. More generally, we recommend the use of cryptographically strong PRNGs in all contexts where randomness is required, as past experience has repeatedly shown that security requirements may arise unexpectedly even in applications that appear uncritical at first.

Prochains exposés

  • Dissecting CRAFT, a full-round attack

    • 18 septembre 2026 (13:45 - 14:45)

    • Batiment 32A salle 15

    Orateur : Eran Lambooij - Inria

    I will present the first full-round key recovery attack on CRAFT, a block cipher introduced at ToSC 2019. The attack builds on the previous observation (ToSC 2026) that the state of CRAFT can be decomposed into two parts that barely exchange information. We transform this property into a dissection attack on the full-round cipher. This shows that in some cases we can elevate the dissection attack[…]
    • Cryptography

  • Key Attack on the ACDGV Matrix Encryption Scheme

    • 25 septembre 2026 (13:45 - 14:45)

    • IRMAR - Université de Rennes - Campus Beaulieu Bat. 22, RDC, Rennes - Amphi Lebesgue

    Orateur : Anmoal Porwal - Technical University of Munich

    I will present our key-recovery attack on the ACDGV public-key encryption scheme proposed at ASIACRYPT 2024 by Aragon, Couvreur, Dyseryn, Gaborit, and Vinçotte. The secret key is a Gabidulin code hidden by appending random rows and columns and by left- and right-multiplication with invertible matrices. Our attack exploits the resulting algebraic structure to recover an equivalent secret key. It[…]
    • Cryptography

    • Asymmetric primitive

  • Module Learning With Errors and Structured Extrapolated Dihedral Cosets

    • 02 octobre 2026 (13:45 - 14:45)

    • IRMAR - Université de Rennes - Campus Beaulieu Bat. 22, RDC, Rennes - Amphi Lebesgue

    Orateur : Jinwei Zheng - Télécom Paris

    The Module Learning With Errors (MLWE) problem is the fundamental hardness assumption underlying the key encapsulation and signature schemes ML-KEM and ML-DSA, which have been selected by NIST for post-quantum cryptography standardization. Understanding its quantum hardness is crucial for assessing the security of these standardized schemes.   Inspired by the equivalence between LWE and[…]
    • Cryptography

Voir les exposés passés