A robust generative classifier against transfer attacks based on variational auto-encoders

Chen Zhang, Zhuo Tang, Youfei Zuo, Kenli Li, Keqin Li

Abstract: Highlights•A robust generative classifier based on variational auto-encoders•The classifier is more robust against transfer-based attacks.•Minimize the KL divergence to get the evidence lower bound•Consider a Gaussian mixture prior on the latent variable.