A Survey of Attacks on Large Vision-Language Models: Resources, Advances, and Future Trends

Daizong Liu,Mingyu Yang,Xiaoye Qu,Pan Zhou,Yu Cheng,Wei Hu

2024-07-12

Abstract:With the significant development of large models in recent years, Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities across a wide range of multimodal understanding and reasoning tasks. Compared to traditional Large Language Models (LLMs), LVLMs present great potential and challenges due to its closer proximity to the multi-resource real-world applications and the complexity of multi-modal processing. However, the vulnerability of LVLMs is relatively underexplored, posing potential security risks in daily usage. In this paper, we provide a comprehensive review of the various forms of existing LVLM attacks. Specifically, we first introduce the background of attacks targeting LVLMs, including the attack preliminary, attack challenges, and attack resources. Then, we systematically review the development of LVLM attack methods, such as adversarial attacks that manipulate model outputs, jailbreak attacks that exploit model vulnerabilities for unauthorized actions, prompt injection attacks that engineer the prompt type and pattern, and data poisoning that affects model training. Finally, we discuss promising research directions in the future. We believe that our survey provides insights into the current landscape of LVLM vulnerabilities, inspiring more researchers to explore and mitigate potential safety issues in LVLM developments. The latest papers on LVLM attacks are continuously collected in <a class="link-external link-https" href="https://github.com/liudaizong/Awesome-LVLM-Attack" rel="external noopener nofollow">this https URL</a>.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper attempts to address the issue of the security and potential vulnerabilities of large vision-language models (LVLMs) in practical applications. Specifically, the paper focuses on the following aspects:

1. **Security Threats of LVLMs**: Although LVLMs perform excellently in multimodal understanding and reasoning tasks, their security has not been fully studied. There are various attack methods, such as adversarial attacks, jailbreak attacks, prompt injection attacks, and data poisoning, which may cause LVLMs to produce incorrect or malicious outputs, or even execute unauthorized operations.

2. **Review of Existing Attack Methods**: The paper provides a comprehensive review of existing LVLM attack methods, including the background, challenges, and resources of the attacks, as well as the development of different types of attack methods. By systematically summarizing these attack methods, the paper aims to provide a clear reference framework for researchers and practitioners.

3. **Future Research Directions**: The paper discusses future research directions to inspire more researchers to explore and mitigate the security issues of LVLMs, thereby improving the robustness and security of these models.

Overall, the goal of the paper is to reveal the current security risks of LVLMs through a comprehensive review of existing LVLM attack methods and to provide guidance for future security research.