Adversarial Attacks
Gokula Krishnan discusses the ongoing cat and mouse game in the literature surrounding adversarial attacks and the potential for future responses. He highlights the robustness of their findings and the interest it has generated among other researchers. Despite concerns about the quality of the gan used, their paper demonstrates the effectiveness of their approach. The chapter also explores the two steps involved in dealing with adversarial examples: identification and cleaning out the adversarial noise.In this clip
From this podcast
Data Skeptic
Defending Against Adversarial Attacks
Related Questions