Provenance-Aware Neutrosophic Governance for Agentic Cybersecurity in 6G Networks: Risk-Constrained Autonomous Response Authorization

Abstract

Agentic AI can support rapid cybersecurity response in highly programmable 6G networks, but autonomous action introduces a critical safety problem: a technically effective response may still disrupt protected services, violate latency constraints, or act on duplicated, stale, correlated, or contradictory evidence. This paper develops a neutrosophic-governed framework for deciding whether an agent-generated cybersecurity response is sufficiently justified for autonomous execution. The central principle is to separate response generation from response authorization. The proposed framework first introduces a provenanceaware signed temporal evidence model that converts heterogeneous cybersecurity observations into a neutrosophic threat state

DOI: 10.5281/zenodo.15825573