Privacy-preserving revocable access control for LLM-driven electrical distributed systems

Abstract

Large Language Models (LLMs) have become transformative tools in natural language processing, significantly advancing the fields of communication, information analysis, and knowledge sharing. However, the vast amounts of sensitive data they handle pose significant challenges for data security and privacy. Traditional cryptographic methods face limitations in efficiently managing access control in LLM-driven electrical distributed systems. This paper introduces a novel Privacy-preserving Revocable Access Control for LLM-driven electrical distributed systems, addressing key concerns such as access policy concealment, user revocation efficiency, and computational overhead. Leveraging an inner-product-based access control mechanism, the proposed scheme achieves complete access policy concealment while supporting flexible access control with wildcard capabilities. Additionally, it facilitates efficient user revocation without requiring costly updates to ciphertexts in such an LLM-driven electrical distributed system. The concise algorithmic structure ensures high efficiency, further enhanced through online/offline encryption and outsourced decryption mechanisms.