Sparse Autoencoders as Plug-and-Play Firewalls for Adversarial Attack Detection in VLMs

TL;DR

This paper introduces SAEgis, a lightweight plug-and-play framework using sparse autoencoders to detect adversarial attacks in vision-language models, enhancing safety without extra adversarial training.

Contribution

It presents the first application of sparse autoencoders as a plug-and-play method for adversarial attack detection in VLMs, improving cross-domain generalization and robustness.

Findings

SAEgis achieves strong detection performance across various attack settings.

Combining multiple layer signals enhances robustness and stability.

The method requires no additional adversarial training and adds minimal overhead.

Abstract

Vision-language models (VLMs) have advanced rapidly and are increasingly deployed in real-world applications, especially with the rise of agent-based systems. However, their safety has received relatively limited attention. Even the latest proprietary and open-weight VLMs remain highly vulnerable to adversarial attacks, leaving downstream applications exposed to significant risks. In this work, we propose a novel and lightweight adversarial attack detection framework based on sparse autoencoders (SAEs), termed SAEgis. By inserting an SAE module into a pretrained VLM and training it with standard reconstruction objectives, we find that the learned sparse latent features naturally capture attack-relevant signals. These features enable reliable classification of whether an input image has been adversarially perturbed, even for previously unseen samples. Extensive experiments show that SAEgis achieves strong performance across in-domain, cross-domain, and cross-attack settings, with particularly large improvements in cross-domain generalization compared to existing baselines. In addition, combining signals from multiple layers further improves robustness and stability. To the best of our knowledge, this is the first work to explore SAE as a plug-and-play mechanism for adversarial attack detection in VLMs. Our method requires no additional adversarial training, introduces minimal overhead, and provides a practical approach for improving the safety of real-world VLM systems.