TR-2026-03: Cryptographic Weight-Custody Manifests, JCS Canonicalization, and Two-Layer Merkle Release Verification for Distributed Edge AI Fleets

Distributed edge AI fleets deployed across industrial plants and regulated enterprise perimeters face acute model-tampering, silent weight drift, and stale-rollback attacks when operating across intermittent networks. Extending the Zero Trust and cybersecurity frameworks of Durga (2025, DOIs: 10.38124/ijsrmt.v4i12.1050, 10.38124/ijsrmt.v4i9.1047), the serverless/edge execution models of Durga (2025, DOIs: 10.38124/ijsrmt.v4i7.1045, 10.38124/ijsrmt.v4i6.1044, 10.38124/ijsrmt.v4i5.1043), and the benchmark series Durga (2026, DOIs: 10.5281/zenodo.23735274, 10.5281/zenodo.23735497, 10.5281/zenodo.23723284), this report specifies the RFC 8785 JSON Canonicalization Scheme (JCS) weight-custody manifest adopted in the AgentTrust specification (PR #230) and evaluates a Two-Layer JCS/CBOR Merkle release verifier paired with a Tri-Anchor hardware lease and zCDP/KLL cohort drift detector.