Images are Achilles' Heel of Alignment: Exploiting Visual Vulnerabilities for Jailbreaking Multimodal Large Language Models

Yifan Li, Hangyu Guo, Kun Zhou, Wayne Xin Zhao, Ji-Rong Wen
2/14/2026

No official code found. Best implementation: AoiDragon/HADES (41 stars, MIT, last commit 2024-10-17). 1 more below.

Abstract

In this paper, we study the harmlessness alignment problem of multimodal large language models (MLLMs). We conduct a systematic empirical analysis of the harmlessness performance of representative MLLMs and reveal that the image input poses the alignment vulnerability of MLLMs. Inspired by this, we propose a novel jailbreak method named HADES, which hides and amplifies the harmfulness of the malicious intent within the text input, using meticulously crafted images. Experimental results show that HADES can effectively jailbreak existing MLLMs, which achieves an average Attack Success Rate (ASR) of 90.26% for LLaVA-1.5 and 71.60% for Gemini Pro Vision. Our code and data are available at https://github.com/RUCAIBox/HADES.

View on arXivView PDF

Code Implementations(2)

RUCAIBox/HADESCommunity100%
MIT

[ECCV'24 Oral] The official GitHub page for ''Images are Achilles' Heel of Alignment: Exploiting Visual Vulnerabilities for Jailbreaking Multimodal Large Language Models''

394CSS, HTMLApr 8, 20241 years agoMIT
AoiDragon/HADESCommunity65%
MIT

[ECCV'24 Oral] The official GitHub page for ''Images are Achilles' Heel of Alignment: Exploiting Visual Vulnerabilities for Jailbreaking Multimodal Large Language Models''

417CSS, HTMLMar 25, 20242 years agoMIT

Similar Papers

Beyond Visual Safety: Jailbreaking Multimodal Large Language Models for Harmful Image Generation via Semantic-Agnostic Inputs

93% match2026

JailBreakV: A Benchmark for Assessing the Robustness of MultiModal Large Language Models against Jailbreak Attacks

90% match2024

ImgTrojan: Jailbreaking Vision-Language Models with ONE Image

90% match1 repo242024

Odysseus: Jailbreaking Commercial Multimodal LLM-integrated Systems via Dual Steganography

90% match1 repo602025

Visual Adversarial Examples Jailbreak Aligned Large Language Models

89% match1 repo2822023

Jailbreaking Multimodal Large Language Models via Shuffle Inconsistency

89% match1 repo162025

Cite this paper

@article{li2026images,
  title  = {Images are Achilles' Heel of Alignment: Exploiting Visual Vulnerabilities for Jailbreaking Multimodal Large Language Models},
  author = {Yifan Li and Hangyu Guo and Kun Zhou and Wayne Xin Zhao and Ji-Rong Wen},
  year   = {2026},
  eprint = {2403.09792},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2403.09792},
  journal = {ECCV 2024 2024}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2403.09792.svg)](https://codeofpaper.com/paper/2403.09792)

Discussion