Adversarial Pruning: A Survey and Benchmark of Pruning Methods for Adversarial Robustness

TL;DR

This paper surveys and benchmarks adversarial pruning methods, proposing a new taxonomy and evaluation framework to fairly compare different approaches and analyze their effectiveness against adversarial attacks.

Contribution

It introduces a novel taxonomy for adversarial pruning methods and a fair benchmark for empirical evaluation, addressing limitations of previous analyses.

Findings

Top-performing methods share common traits

Many methods face similar limitations

Benchmark results highlight key differences

Abstract

Recent work has proposed neural network pruning techniques to reduce the size of a network while preserving robustness against adversarial examples, i.e., well-crafted inputs inducing a misclassification. These methods, which we refer to as adversarial pruning methods, involve complex and articulated designs, making it difficult to analyze the differences and establish a fair and accurate comparison. In this work, we overcome these issues by surveying current adversarial pruning methods and proposing a novel taxonomy to categorize them based on two main dimensions: the pipeline, defining when to prune; and the specifics, defining how to prune. We then highlight the limitations of current empirical analyses and propose a novel, fair evaluation benchmark to address them. We finally conduct an empirical re-evaluation of current adversarial pruning methods and discuss the results, highlighting the shared traits of top-performing adversarial pruning methods, as well as common issues. We welcome contributions in our publicly-available benchmark at https://github.com/pralab/AdversarialPruningBenchmark