Franck Leprévost

Prof. Dr. Franck Leprévost

orcid

Research Focus & Metrics

Academic Career Length

35Years (1992-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 December 1992 in manuscripta mathematica
Verified
Journal Article
Closed access
PDF file
FLFranck Leprévost
Franck Leprévost

10 October 2020 in Journal of Information and Telecommunication
Verified
Journal Article
Open access
RCRaluca Chitic
Raluca Chitic
FLFranck Leprévost
Franck Leprévost
NBNicolas Bernard
Nicolas Bernard
The range of applications of Neural Networks encompasses image classification. However, Neural Networks are vulnerable to attacks, and may misclassify adversarial images, leading to potentially disastrous consequences. Pursuing some of our previous work, we provide an extended proof of concept of a black-box, targeted, non-parametric attack using evolutionary algorithms to fool both Neural Networks and humans at the task of image classification. Our feasibility study is performed on VGG-16 trained on CIFAR-10. For any category cA of CIFAR-10, one chooses an image A classified by VGG-16 as belonging to cA. From there, two scenarios are addressed. In the first scenario, a target category ct≠cA is fixed a priori. We construct an evolutionary algorithm that evolves A to a modified image that VGG-16 classifies as belonging to ct. In the second scenario, we construct another evolutionary algorithm that evolves A to a modified image that VGG-16 is unable to classify. In both scenarios, the obtained adversarial images remain so close to the original one that a human would likely classify them as still belonging to A.



14 September 2023 in IEEE Access
Verified
Journal Article
Open access
BPBernardo Pulido-Gaytan
Bernardo Pulido-Gaytan
Andrei TchernykhAndrei Tchernykh
Andrei Tchernykh
FLFranck Leprévost
Franck Leprévost
Pascal BouvryPascal Bouvry
Pascal Bouvry
AGAlfredo Goldman
Alfredo Goldman
The security issues that arise in public cloud environments raise several concerns about privacy-preserving. Conventional security practices successfully protect stored and transmitted data by encryption, but not during data processing where the data value extraction requires decryption. It creates critical exposure points for sensitive sectors like healthcare, pharmaceutical, genomics, government, and financial, among many others that cause hesitation to use these third-party services and prevent widespread practical adoption of cloud solutions. Homomorphic Encryption (HE) emerges as a mechanism for expanding the scope of public cloud services for sensitive data processing. However, high-demand solutions such as artificial intelligence and machine learning require efficient operations beyond HE additions and multiplications. In this paper, we analyze the current homomorphic comparison methods across their strengths and weaknesses and present theoretical concepts, state-of-the-art techniques, challenges, opportunities, and open problems. We theoretically prove the limits of the representability of sign and comparison functions in polynomial forms for HE schemes. We show that both functions can be represented as polynomials over the Galois field and cannot be represented over a residue ring with zero divisors. We compare the efficiency, accuracy, and computational complexity of different homomorphic comparison approaches. The experimental results demonstrate that Newton-Raphson is the fastest method for generating polynomial approximations and evaluating comparisons, and the Fourier method is the most accurate considering the L 1 , L 2 , L ∞ norms and R 2 measure. The bi-objective analysis presents the performance compromise between complexity and accuracy.

4 December 2020
Verified
Book Chapter
Closed access
PDF file
JCJorge M. Cortés-Mendoza
Jorge M. Cortés-Mendoza
Andrei TchernykhAndrei Tchernykh
Andrei Tchernykh
Mikhail BabenkoMikhail Babenko
Mikhail Babenko
LPLuis Bernardo Pulido-Gaytán
Luis Bernardo Pulido-Gaytán
GRGleb Radchenko
Gleb Radchenko
AAArutyun Avetisyan
Arutyun Avetisyan