Agentic AI for Self-Sovereign Identity: A Decentralized Zero Trust Framework for Autonomous Microservices

The rapid evolution of Agentic Artificial Intelligence (AI)—autonomous, context-aware agents capable of self-directed decision-making—has introduced unprecedented security challenges for microservices architectures. Traditional session-based authentication, dependent on static tokens and centralized identity providers, is ill-suited for the dynamic, ephemeral, and machine-to-machine (M2M) interactions prevalent in zero trust environments. This paper investigates the convergence of Agentic AI and decentralized identity (DID) frameworks, emphasizing the role of verifiable credentials (VCs), dynamic token issuance, and contextual access control in enabling scalable, trust-minimized (i.e., reducing reliance on centralized authorities) service interactions. We propose a decentralized authentication and authorization framework where DIDs, maintained on blockchain-based registries, replace conventional identity silos, enabling autonomous agents to cryptographically prove trustworthiness without relying on persistent session states. Context-aware policy engines evaluate real-time telemetry such as location, workload, and behavioural patterns to issue short-lived, ephemeral access tokens with adaptive time-to-live (TTL) values. Experimental results from a Kubernetes-based microservices testbed with 50 simulated agents show that the proposed approach reduces authentication latency by 50% (from 180 ms to 90 ms), eliminates token replay vulnerabilities, and increases authentication throughput by 75% (from 800 to 1,400 agents/min) compared to OAuth2/JWT baselines. Furthermore, dynamic policy adaptation ensures immediate revocation of access when agents deviate from expected operational norms, minimizing attack surfaces. This work offers a novel synthesis of AI autonomy and decentralized identity principles, delivering both performance gains and enhanced security in zero trust microservices. The proposed architecture paves the way for resilient, self-governing ecosystems where Agentic AI can operate securely, efficiently, and adaptively in highly dynamic environments.