Denoised Internal Models: a Brain-Inspired Autoencoder against Adversarial Attacks

Kaiyuan Liu, Xingyu Li, Yi Zhou, Jisong Guan, Yurui Lai, Hang Su, Ge Zhang, Jiachen Wang, Chunxu Guo

Keywords: adversarial attacks, robustness of neural network, autoencoder, visual cortex, engram cell

Abstract: Despite its great success, deep learning severely suffers from robustness; that is, deep neural networks are very vulnerable to adversarial attacks, even the simplest ones. Inspired by recent advances in brain science, we propose the Denoised Internal Models (DIM), a novel generative autoencoder-based model to tackle this challenge. Simulating the pipeline in the human brain for visual signal processing, DIM adopts a two-stage approach. In the first stage, DIM uses a denoiser to reduce the noise and the dimensions of inputs, reflecting the information pre-processing in the thalamus. Inspired from the sparse coding of memory-related traces in the primary visual cortex, the second stage produces a set of internal models, one for each category. We evaluate DIM over 42 adversarial attacks, showing that DIM effectively defenses against all the attacks and outperforms the SOTA on the overall robustness.

One-sentence Summary: Inspired by the visual process system of the brain, we propose the Denoised Internal Models as an overall robust model defensed on all the attacks.