advertorch v0.1: An Adversarial Robustness Toolbox based on PyTorch

TL;DR

The paper introduces advertorch v0.1, a PyTorch-based toolbox for adversarial robustness research spanning attacks, defenses, and robust training. It emphasizes simple, consistent APIs and concise reference implementations suitable for attack-in-the-loop workflows, leveraging dynamic computational graphs for speed. It documents a wide range of gradient-based and other attacks, a BPDA wrapper for non-differentiable defenses, and preprocessing defenses plus robust training references. It also outlines a versioning scheme and reproducible benchmark reporting to support community collaboration and benchmarking.

Abstract

advertorch is a toolbox for adversarial robustness research. It contains various implementations for attacks, defenses and robust training methods. advertorch is built on PyTorch (Paszke et al., 2017), and leverages the advantages of the dynamic computational graph to provide concise and efficient reference implementations. The code is licensed under the LGPL license and is open sourced at https://github.com/BorealisAI/advertorch .