Decentralized Digital Identity Management for Large Language Model Agents

A Large Language Model (LLM) agent is an AI system that uses an LLM as its core to interact with its environment and perform tasks that go beyond text generation by integrating reasoning, planning, and external tools. While these agents automate complex language-based tasks, they face security risks such as prompt injection attacks, unauthorized access, and exploitation by external Application Programming Interfaces (APIs). This paper explores integrating LLMs with blockchain-based Self-Sovereign Identity (SSI) systems to enhance digital governance in AI applications. Experimental results show that integration can improve security, achieving up to 100% validation success at low load and 93% at peak load surpassing traditional authentication systems. However, improved security comes with an increased cost in terms of latency and throughput, which requires further investigation into optimizations such as Layer 2 solutions for scalability.