Translating Natural Language Specifications into Access Control Policies by Leveraging Large Language Models
Sherifdeen Lawal, Xingmeng Zhao, Anthony Rios, Ram Krishnan, David Ferraiolo
Abstract
This paper investigates the application of large language models (LLMs) for the automated translation and information extraction of access control policies from a natural language source. Prior research in this domain have predominantly relied on manual methods, traditional natural language processing (NLP), or a hybrid approach involving machine learning and artificial neural networks combined with NLP techniques. We demonstrate a significant advancement by leveraging the power of LLMs to achieve improved efficiency and accuracy in these tasks. Our study focuses on applying cutting-edge prompt engineering techniques designed to optimize LLM performance in the specific context of access control policy information extraction. The findings highlight the potential of LLMs to streamline the process of converting human-readable requirements into formal, machine-interpretable policies, ultimately contributing to the automation and security of access control systems.