[2405.04760] Large Language Models for Cyber Security: A Systematic Literature Review

View the PDF file for the paper entitled Large Language Models for Security

PDF HTML (experimental) view

a summary:The rapid progress of large language models (LLMS) has opened new opportunities to take advantage of artificial intelligence in various fields, including cybersecurity. With the growth and development of cyber threats, there is an increasing need for smart systems that can discover weaknesses automatically, analyze harmful programs and respond to attacks. In this poll, we are conducting a comprehensive review of literature on the LLMS application in the LLM4security. By collecting more than 30 thousand paper related to a comprehensive way and analyzing 127 papers from the best places of security and software engineering, we aim to provide a comprehensive vision of how LLMS is used to solve various problems through the field of cybersecurity. By analyzing, we define many major results. First, we note that LLMS is applied to a wide range of cybersecurity, including weaknesses, malware analysis, network infiltration discovery, and clinic detection. Second, we find that the data sets used for training and evaluation of LLMS in these tasks are often limited in size and diversity, highlighting the need for more comprehensive and representative data groups. Third, we define many promising techniques for LLMS air conditioning with specific cybersecurity, such as installation, transport learning, and field training. Finally, we discuss the main challenges and opportunities for future research in Llm4security, including the need for more interpretable and interpretable models, the importance of addressing the privacy of data and security concerns, and the ability to benefit from LLMS for pre -emptive defense and hunting threat. In general, our survey provides a comprehensive overview of the latest latest LLM4security and defines many promising trends for future research.