# Learning Adversarially Robust Representations via Worst-Case Mutual Information Maximization

arXiv: https://arxiv.org/abs/2002.11798 · HTML: https://mlsanity.com/paper/2002.11798/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Feb 2020.

## Built on by (most cited first, 6 of 6)

- [Better Safe Than Sorry: Preventing Delusive Adversaries with Adversarial Training](https://mlsanity.com/paper/2102.04716/) (2021)
- [Adversarial Examples can be Effective Data Augmentation for Unsupervised Machine Learning](https://mlsanity.com/paper/2103.01895/) (2021)
- [Improving Adversarial Robustness via Mutual Information Estimation](https://mlsanity.com/paper/2207.12203/) (2022)
- [Toward Enhanced Robustness in Unsupervised Graph Representation Learning: A Graph Information Bottleneck Perspective](https://mlsanity.com/paper/2201.08557/) (2022)
- [Bayesian Learning with Information Gain Provably Bounds Risk for a Robust Adversarial Defense](https://mlsanity.com/paper/2212.02003/) (2022)
- [Measure-Theoretic Anti-Causal Representation Learning](https://mlsanity.com/paper/2510.18052/) (2025)

## Abstract (arXiv)

Training machine learning models that are robust against adversarial inputs poses seemingly insurmountable challenges. To better understand adversarial robustness, we consider the underlying problem of learning robust representations. We develop a notion of representation vulnerability that captures the maximum change of mutual information between the input and output distributions, under the worst-case input perturbation. Then, we prove a theorem that establishes a lower bound on the minimum adversarial risk that can be achieved for any downstream classifier based on its representation vulnerability. We propose an unsupervised learning method for obtaining intrinsically robust representations by maximizing the worst-case mutual information between the input and output distributions. Experiments on downstream classification tasks support the robustness of the representations found using unsupervised learning with our training principle.