Exploiting Vision-LLM Vulnerability: Enhancing Typographic Attacks with Instructional Directives

Table of Links

Abstract and 1. Introduction

2. Related Work

   2.1 Vision-LLMs

   2.2 Transferable Adversarial Attacks

3. Preliminaries

   3.1 Revisiting Auto-Regressive Vision-LLMs

   3.2 Typographic Attacks in Vision-LLMs-based AD Systems

4. Methodology

   4.1 Auto-Generation of Typographic Attack

   4.2 Augmentations of Typographic Attack

   4.3 Realizations of Typographic Attacks

5. Experiments

6. Conclusion and References

4.2 Augmentations of Typographic Attack

Inspired by the success of instruction-prompting methodologies [37, 38], the greedy reasoning in LLMs [39], and to further exploit the ambiguity between textual and visual tokens in Vision-LLMs, we propose to augment the typographic attacks prompts within images by explicitly providing instruction keywords that emphasize text-to-text alignment over that of visual-language tokens. Our approach realizes the concept in the form of instructional directives: ❶ command directives for emphasizing a false answer and ❷ conjunction directives to additionally include attack clauses. In particular, we have developed,

• Command Directive. By embedding commands with the attacks, we aim to prompt the VisionLLMs into greedily producing erroneous answers. Our work investigates the "ANSWER:" directive as a prefix before the first attack prompt.

• Conjunction Directive. Conjunctions, connectors (or the lack thereof) act to link together separate attack concepts that make the overall text appear more coherent, thereby increasing the likelihood of multi-task success. In our work, we investigate these directives as "AND," "OR," "WITH," or simply empty spaces as prefixes between attack prompts.

While other forms of directives can also be useful for enhancing the attack success rate, we focus on investigating basic directives related to typographic attacks in this work.