Every Picture Tells a Dangerous Story: Memory-Augmented Multi-Agent Jailbreak Attacks on VLMs

TL;DR

This paper introduces MemJack, a memory-augmented multi-agent framework that exploits visual semantic vulnerabilities to perform effective jailbreak attacks on vision-language models, revealing significant security risks.

Contribution

MemJack is the first to leverage visual semantics and multi-agent cooperation for automated, persistent jailbreak attacks on VLMs, significantly improving attack success rates.

Findings

MemJack achieves 71.48% attack success rate on COCO images.

Success rate scales to 90% with extended attack budgets.

A new dataset with over 113,000 attack trajectories will be released.

Abstract

The rapid evolution of Vision-Language Models (VLMs) has catalyzed unprecedented capabilities in artificial intelligence; however, this continuous modal expansion has inadvertently exposed a vastly broadened and unconstrained adversarial attack surface. Current multimodal jailbreak strategies primarily focus on surface-level pixel perturbations and typographic attacks or harmful images; however, they fail to engage with the complex semantic structures intrinsic to visual data. This leaves the vast semantic attack surface of original, natural images largely unscrutinized. Driven by the need to expose these deep-seated semantic vulnerabilities, we introduce \textbf{MemJack}, a \textbf{MEM}ory-augmented multi-agent \textbf{JA}ilbreak atta\textbf{CK} framework that explicitly leverages visual semantics to orchestrate automated jailbreak attacks. MemJack employs coordinated multi-agent cooperation to dynamically map visual entities to malicious intents, generate adversarial prompts via multi-angle visual-semantic camouflage, and utilize an Iterative Nullspace Projection (INLP) geometric filter to bypass premature latent space refusals. By accumulating and transferring successful strategies through a persistent Multimodal Experience Memory, MemJack maintains highly coherent extended multi-turn jailbreak attack interactions across different images, thereby improving the attack success rate (ASR) on new images. Extensive empirical evaluations across full, unmodified COCO val2017 images demonstrate that MemJack achieves a 71.48\% ASR against Qwen3-VL-Plus, scaling to 90\% under extended budgets. Furthermore, to catalyze future defensive alignment research, we will release \textbf{MemJack-Bench}, a comprehensive dataset comprising over 113,000 interactive multimodal jailbreak attack trajectories, establishing a vital foundation for developing inherently robust VLMs.