Bypassing Backdoor Detection Algorithms in Deep Learning

Te Juin Lester Tan, Reza Shokri

Deep learning models are vulnerable to various adversarial manipulations of their training data, parameters, and input sample. In particular, an adversary can modify the training d…