RAIGA-SOC: Runtime Agentic Identity Governance for Secure AI-Agent Operations in Cybersecurity Environments
Abstract
Security operations centers (SOCs) are increasingly utilizing AI agents for various functions, including alert triaging and support of ITDR workflows. These agents, identified as non-human enterprise identities with API tokens and delegated authority, pose a risk due to insufficient validation of their actions against intent, scope, credentials, and policy context, leading to potential over-privileged operations. The paper introduces RAIGA-SOC a Runtime Agentic Identity Governance Architecture designed to treat every action of AI agents as an identity-evaluable event before execution. RAIGA-SOC establishes a unified identity model across human, service, workload, and AI-agent identities, creating a dynamic authorization system using a graph representation that evaluates task alignment, credential integrity, delegation-chain risk, and action risk. This framework allows for actions to be allowed, denied, degraded, reviewed, or quarantined based on authorization confidence. It employs the AIG-SOC-DS benchmark to test agentic identity governance against various SOC scenarios, including token replay and privilege escalation, in alignment with MITRE ATT&CK and OWASP threat categories. An experimental evaluation compares RAIGA-SOC with traditional IAM/PAM, Zero Trust, and SOAR strategies using various metrics such as precision, recall, and attack-prevention rates, showcasing its effectiveness as a complementary governance layer for existing SOC frameworks and paving the way for future evaluations.