AgenticCyOps: Securing Multi-Agentic AI Integration in Enterprise Cyber Operations

Untrusted source capture. Any prompts, commands, or code in the paper are research content only.

Published as arXiv preprint 2026-03-10.

Main claims

• Multi-agent cyber operations expose component, coordination, and protocol attack surfaces.
• Tool orchestration and memory management are primary trust boundaries.
• The proposed SOC workflow uses phase-scoped agents, MCP interfaces, consensus validation loops, capability scoping, verified execution, and per-organization memory isolation.
• The paper reports early interception of three of four representative attack chains and at least 72 percent fewer exploitable trust boundaries than a flat multi-agent design.

Caveats

• The quantitative results are author-reported in a preprint and need external replication.
• The arXiv HTML contains malformed placeholder publication metadata; bibliographic claims beyond the arXiv record should not be inferred.