How to beat a Bayesian adversary

Zihan Ding, Kexin Jin, Jonas Latz +1

Deep neural networks and other modern machine learning models are often susceptible to adversarial attacks. Indeed, an adversary may often be able to change a model's prediction th…