Revisiting Hilbert-Schmidt Information Bottleneck for Adversarial Robustness
This work addresses the problem of adversarial vulnerability in machine learning models, offering an incremental improvement through a novel regularization approach.
The paper tackles improving adversarial robustness in deep neural networks by using the Hilbert-Schmidt independence criterion (HSIC) bottleneck as a regularizer, showing that it reduces sensitivity to adversarial attacks and achieves competitive natural accuracy with enhanced robustness on benchmark datasets.
We investigate the HSIC (Hilbert-Schmidt independence criterion) bottleneck as a regularizer for learning an adversarially robust deep neural network classifier. In addition to the usual cross-entropy loss, we add regularization terms for every intermediate layer to ensure that the latent representations retain useful information for output prediction while reducing redundant information. We show that the HSIC bottleneck enhances robustness to adversarial attacks both theoretically and experimentally. In particular, we prove that the HSIC bottleneck regularizer reduces the sensitivity of the classifier to adversarial examples. Our experiments on multiple benchmark datasets and architectures demonstrate that incorporating an HSIC bottleneck regularizer attains competitive natural accuracy and improves adversarial robustness, both with and without adversarial examples during training. Our code and adversarially robust models are publicly available.