Adversarial attacks against Modern Vision-Language Models

TL;DR

This paper evaluates the adversarial robustness of open-source vision-language models in a simulated e-commerce environment, revealing significant vulnerabilities in some models and highlighting the importance of security assessments before deployment.

Contribution

It provides the first comprehensive analysis of gradient-based adversarial attacks on open-source VLM agents in a realistic setting, comparing their robustness.

Findings

LLaVA-v1.5-7B is vulnerable to gradient-based attacks with over 50% success rate.

Qwen2.5-VL-7B shows higher robustness with less than 16% success rate.

Architectural differences influence the adversarial resilience of VLMs.

Abstract

We study adversarial robustness of open-source vision-language model (VLM) agents deployed in a self-contained e-commerce environment built to simulate realistic pre-deployment conditions. We evaluate two agents, LLaVA-v1.5-7B and Qwen2.5-VL-7B, under three gradient-based attacks: the Basic Iterative Method (BIM), Projected Gradient Descent (PGD), and a CLIP-based spectral attack. Against LLaVA, all three attacks achieve substantial attack success rates (52.6%, 53.8%, and 66.9% respectively), demonstrating that simple gradient-based methods pose a practical threat to open-source VLM agents. Qwen2.5-VL proves significantly more robust across all attacks (6.5%, 7.7%, and 15.5%), suggesting meaningful architectural differences in adversarial resilience between open-source VLM families. These findings have direct implications for the security evaluation of VLM agents prior to commercial deployment.