- A.I. seminar: Challenges of A.I.: integrity ,confidentiality and availability
WINTER 2022-2023 & WINTER 2023-2024
Given to: second year master students of the Université Côte d'Azur (master degree: mathematics for engineers, specialization in machine learning).
Overview of main threats targeting a machine learning system under the CIA (Confidentiality, Integrity, Availability) triad, with examples for each of them : poisoning attacks, membership inference attacks, reverse engineering attacks, adversarial examples.
Theoretical presentation of adversarial examples: intuition, reasons for their existence, technical details of attacks and defense schemes in different settings (white-box, black-box).
- Introduction to attacks against a machine learning based system
WINTER 2020-2021 & WINTER 2021-2022
Given to: senior students of the engineering school "Ecole des Mines de Saint-Etienne".
Overview of main threats targeting a machine learning system under the CIA (Confidentiality, Integrity, Availability) triad, with examples for each of them : poisoning attacks, membership inference attacks, reverse engineering attacks, adversarial examples.
poisoning attacks, membership inference attacks, reverse engineering attacks, adversarial examples.
Presentation of basic gradient-based and score-based attacks: FGSM, BIM, CWL2, ZOO and SPSA.
Presentation of adversarial training and some of its variants to defend against adversarial examples.
Presentation of Differential Privacy to defend against membership inference attacks.
- Algorithms and programming
WINTER 2021-2022
Given to: junior students of the engineering school "Ecole des Mines de Saint-Etienne".
Study and implementation of graph theory algorithms in C: Dijkstra, Bellman-Ford and Chandy-Misra-Haas.