paper-with-me

홈 › Papers

Universal Adversarial Attack on Aligned Multimodal LLMs

2025-02-11 · Temurbek Rahmatullaev, Polina Druzhinina, Matvey Mikhalchuk, Andrey Kuznetsov, Anton Razzhigaev

We propose a universal adversarial attack on multimodal Large Language Models (LLMs) that leverages a single optimized image to override alignment safeguards across diverse queries and even multiple models. By backpropagating through the vision encoder and language head, we craft a synthetic image that forces the model to respond with a targeted phrase (e.g., ''Sure, here it is'') or otherwise unsafe content-even for harmful prompts. In experiments on the SafeBench benchmark, our method achieves significantly higher attack success rates than existing baselines, including text-only universal prompts (e.g., up to 93% on certain models). We further demonstrate cross-model transferability by training on several multimodal LLMs simultaneously and testing on unseen architectures. Additionally, a multi-answer variant of our approach produces more natural-sounding (yet still malicious) responses. These findings underscore critical vulnerabilities in current multimodal alignment and call for more robust adversarial defenses. We will release code and datasets under the Apache-2.0 license. Warning: some content generated by Multimodal LLMs in this paper may be offensive to some readers.

📄 PDF Abstract BibTeX arXiv:2502.07987

Code (0)

등록된 구현이 없습니다.

Tasks

Adversarial Attack

Similar Papers 제목 키워드 기반

Visual Adversarial Examples Jailbreak Aligned Large Language Models

2023-06-22 · Xiangyu Qi, Kaixuan Huang, Ashwinee Panda, Peter Henderson 외

Recently, there has been a surge of interest in integrating vision into Large Language Models (LLMs), exemplified by Visual Language Models (VLMs) such as Flamingo and GPT-4. This paper sheds light on the security and sa…

AmpleGCG: Learning a Universal and Transferable Generative Model of Adversarial Suffixes for Jailbreaking Both Open and Closed LLMs

2024-04-11 · Zeyi Liao, Huan Sun

As large language models (LLMs) become increasingly prevalent and integrated into autonomous systems, ensuring their safety is imperative. Despite significant strides toward safety alignment, recent work GCG~\citep{zou20…

Safety Alignment

Align is not Enough: Multimodal Universal Jailbreak Attack against Multimodal Large Language Models

2025-06-02 · Youze Wang, WenBo Hu, Yinpeng Dong, Jing Liu 외

Large Language Models (LLMs) have evolved into Multimodal Large Language Models (MLLMs), significantly enhancing their capabilities by integrating visual information and other types, thus aligning more closely with the n…

Safety Alignment

Enhancing Adversarial Attacks through Chain of Thought

2024-10-29 · Jingbo Su

Large language models (LLMs) have demonstrated impressive performance across various domains but remain susceptible to safety concerns. Prior research indicates that gradient-based adversarial attacks are particularly ef…

Universal Adversarial Attacks against Closed-Source MLLMs via Target-View Routed Meta Optimization

2026-01-30 · Hui Lu, Yi Yu, Yiming Yang, Chenyu Yi 외 arxiv

Targeted adversarial attacks on closed-source multimodal large language models (MLLMs) have been increasingly explored under black-box transfer, yet prior methods are predominantly sample-specific and offer limited reusa…