1 paper · 1 filter
Siqi Li, Yasser Shoukry
Deep neural networks are highly susceptible to adversarial attacks, which pose significant risks to security- and safety-critical applications. We present KoALA (KL-L0 Adversarial…