Adversarial Attacks on Large Language Models

Large Language Models (LLMs) have rapidly advanced and garnered increasing attention due to their remarkable capabilities across various applications. However, adversarial attacks pose a significant threat to LLMs, as prior research has demonstrated their vulnerability, resulting in prediction inaccuracies. This paper offers a foundational overview of LLMs and traces their developmental trajectory. We systematically classify and compare adversarial examples on LLMs based on their perturbation units. Additionally, we scrutinize the root causes of vulnerability and explore prevalent defense approaches tailored to mitigate adversarial attacks on LLMs.