A Little Robustness Goes a Long Way: Leveraging Robust Features for Targeted Transfer Attacks
This work addresses a key limitation in adversarial machine learning for security applications by enhancing targeted attack transferability, though it is incremental as it builds on prior optimization-focused methods.
The paper tackled the problem of low transferability in targeted adversarial attacks between different neural network architectures by showing that training the source classifier to be slightly robust to small adversarial perturbations substantially improves transferability, achieving up to 40% success rates in cross-architecture attacks.
Adversarial examples for neural network image classifiers are known to be transferable: examples optimized to be misclassified by a source classifier are often misclassified as well by classifiers with different architectures. However, targeted adversarial examples -- optimized to be classified as a chosen target class -- tend to be less transferable between architectures. While prior research on constructing transferable targeted attacks has focused on improving the optimization procedure, in this work we examine the role of the source classifier. Here, we show that training the source classifier to be "slightly robust" -- that is, robust to small-magnitude adversarial examples -- substantially improves the transferability of class-targeted and representation-targeted adversarial attacks, even between architectures as different as convolutional neural networks and transformers. The results we present provide insight into the nature of adversarial examples as well as the mechanisms underlying so-called "robust" classifiers.