Adversarial Attacks Neutralization via Data Set Randomization

Mouna Rabhi, Roberto Di Pietro

Adversarial attacks on deep-learning models pose a serious threat to their reliability and security. Existing defense mechanisms are narrow addressing a specific type of attack or…