Training Robust Neural Networks under Adversarial Attacks

Project for the IASD Master program between Paris-Dauphine, École Normale Supérieure, and Mines ParisTech.

Check the Jupyter Notebook: robust_deep_net

Link to the project presentation slides.

References:

• Goodfellow, I.J., Shlens, J., & Szegedy, C. (2015). Explaining and Harnessing Adversarial Examples. [PDF], [arXiv].
• Madry, A., Makelov, A., Schmidt, L., Tsipras, D., & Vladu, A. (2018). Towards Deep Learning Models Resistant to Adversarial Attacks. [PDF], [arXiv].
• Ilyas, A., Engstrom, L., Athalye, A., & Lin, J. (2018). Black-box Adversarial Attacks with Limited Queries and Information. [PDF], [arXiv].