Description
To understand the sensitivity under attacks and to develop defense mechanisms, machine-learning model designers craft worst-case adversarial perturbations with gradient-descent optimization algorithms against the model under evaluation. However, many of the proposed defenses have been shown to provide a false sense of robustness due to failures of the attacks, rather than actual improvements in the machine‐learning models’ robustness, as highlighted by more rigorous evaluations. Although guidelines and best practices have been suggested to improve current adversarial robustness evaluations, the lack of automatic testing and debugging tools makes it difficult to apply these recommendations in a systematic and automated manner. To this end, the analysis of failures in the optimization of adversarial attacks is the only valid strategy to avoid repeating mistakes of the past.
Next sessions
-
The Battle Against Bots: Current Threats and New Directions to Counter Automated Attacks
Speaker : Elisa Chiapponi - Amadeus IT Group
In today's digital landscape, the battle between industry and automated bots is an ever-evolving challenge. Attackers are leveraging advanced techniques such as residential proxies, CAPTCHA farms, and AI-enhanced fingerprint rotations to evade detection and execute functional abuse attacks, including web scraping, denial of inventory, and SMS pumping. This talk will explore ongoing efforts[…]-
SoSysec
-
Intrusion detection
-
-
Safety-Security Convergence of Industrial Control Systems
Speaker : Maxime Puys - Université Clermont Auvergne - IUT de Clermont-Ferrand
Industrial Control Systems (ICS) are designed to provide a service, such as power generation or water treatment, while protecting people, assets, and the environment against hazards. However, ICS now integrate Information Technology (IT) and are interconnected with the outside world such as the Internet, thereby exposing their infrastructures to cyberattacks. Cyberattacks have thus become new[…]-
SoSysec
-
Intrusion detection
-