Security and Privacy Challenges of Large Language Models: A Survey
Badhan Chandra Das Knight Foundation School of Computing and Information Sciences; Sustainability, Optimization, and Learning for InterDependent networks laboratory (solid lab), Florida International UniversityMiamiFloridaUnited States , M. Hadi Amini Knight Foundation School of Computing and Information Science, solid lab, Florida International UniversityMiamiFloridaUnited States and Yanzhao Wu Knight Foundation School of Computing and Information Sciences, Florida International UniversityMiamiFloridaUnited States Emails:
Abstract
Large language models (LLMs) have demonstrated extraordinary capabilities and contributed to multiple fields, such as generating and summarizing text, language translation, and question-answering. Nowadays, LLMs have become very popular tool in natural language processing (NLP) tasks, with the capability to analyze complicated linguistic patterns and provide relevant and appropriate responses depending on the context. While offering significant advantages, these models are also vulnerable to security and privacy attacks, such as jailbreaking attacks, data poisoning attacks, and personally identifiable information (PII) leakage attacks. This survey provides a thorough review of the security and privacy challenges of LLMs, along with the application-based risks in various domains, such as transportation, education, and healthcare. We assess the extent of LLM vulnerabilities, investigate emerging security and privacy attacks for LLMs, and review the potential defense mechanisms. Additionally, the survey outlines existing research gaps in this research area and highlights future research directions.
中文速览
大型语言模型（LLM）在文本生成、翻译、问答等任务上表现出色，但随之而来的安全与隐私风险却尚未得到系统性梳理。这篇综述针对这一空白，全面整理了LLM面临的主要威胁，包括越狱攻击（jailbreaking）、后门攻击（backdoor attack）、数据投毒（data poisoning）以及个人身份信息（PII）泄露等，同时介绍了各类防御机制，并延伸分析了LLM在交通、教育、医疗等具体应用场景中的特有风险。研究发现，LLM的Transformer架构、基于人类反馈的强化学习（RLHF）和上下文学习等核心机制在赋予模型强大能力的同时，也引入了多种可被攻击者利用的漏洞，而现有防御手段在应对新型攻击时仍存在明显不足。这项工作为研究者和从业者提供了一份清晰的全景图，有助于推动LLM向更安全、更可信的方向发展，在医疗、金融等对隐私和安全高度敏感的领域尤具现实意义。
原文 arXiv:2402.00888;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2402.00888v2