AgenticCyOps: 72% Reduction in Exploitable Trust Boundaries for Enterprise Multi-Agent Security

arXiv·medium signal

AgenticCyOps maps the attack surface of multi-agent LLM systems across component, coordination, and protocol layers, identifying tool orchestration and memory management as primary MCP integration vulnerabilities. Applied to a Security Operations Center workflow, the proposed defensive framework achieves a minimum 72% reduction in exploitable trust boundaries and successfully intercepts 3 of 4 representative attack chains early in execution. The framework aligns with NIST and ISO 27001 standards.