State-Dependent Representation Geometry for Cybersecurity: Authenticated Causal Representation Rotation Under Telemetry- Basis Churn

This paper introduces State-Dependent Representation Geometry (SDRG) as a cybersecurity framework for systems whose telemetry, feature basis, protocol representation, identity structure, or operational topology changes over time. The central mechanism, Authenticated Causal Representation Rotation (CRR), allows a defender to regenerate the active representation geometry from trusted causal history, recover observations into a canonical analytic space, and verify security invariants before detection or control decisions are made. The paper formalizes SDRG through admissible geometry generators, reversible transports, causal synchronization, invariant gates, authenticated transition records, and fail-safe fallback behavior. It positions SDRG as an adaptive layer complementing established cryptography, zero trust, intrusion detection, telecom security, post-quantum migration, and cyber-resilience mechanisms rather than replacing them. Evaluation is performed through CRR-SOTA-4, a telemetry-basis-churn benchmark constructed from four established cybersecurity datasets: UNSW-NB15, CICIDS2017 Friday-DDoS, PhiUSIIL, and NATICUSdroid. Across five adversarial representation-churn scenarios and eight causal-history seeds per dataset, SDRG-CRR records 160 wins in 160 matched paired evaluations against the strongest evaluated non-SDRG baseline, with zero recorded hard-invariant violations and near-exact inverse reconstruction. The reported result establishes the best evaluated performance on CRR-SOTA-4 v1.0 for authenticated causal representation recovery under telemetry churn. It is not presented as a universal cybersecurity or clean-dataset leaderboard claim. The accompanying reproducibility bundle contains benchmark code, configuration, paired results, audits, and supporting artifacts.