Red Teaming the Mind of the Machine: A Systematic Evaluation of Prompt Injection and Jailbreak Vulnerabilities in LLMs
This addresses security vulnerabilities in LLMs for developers and users, though it is incremental as it builds on existing red-teaming approaches.
The paper systematically investigates jailbreak strategies against state-of-the-art LLMs, categorizing over 1,400 adversarial prompts and analyzing their success across models like GPT-4 and Claude 2, while proposing layered mitigation strategies for enhanced security.
Large Language Models (LLMs) are increasingly integrated into consumer and enterprise applications. Despite their capabilities, they remain susceptible to adversarial attacks such as prompt injection and jailbreaks that override alignment safeguards. This paper provides a systematic investigation of jailbreak strategies against various state-of-the-art LLMs. We categorize over 1,400 adversarial prompts, analyze their success against GPT-4, Claude 2, Mistral 7B, and Vicuna, and examine their generalizability and construction logic. We further propose layered mitigation strategies and recommend a hybrid red-teaming and sandboxing approach for robust LLM security.