BAN: Detecting Backdoors Activated by Adversarial Neuron Noise

Neural Information Processing Systems 

Backdoor attacks on deep learning represent a recent threat that has gained significant attention in the research community.