DEOS Trust Substrate — Consolidated Threat Model v1

Status: v1.1 — drafted 2026-05-05 Spec URI: https://projectmaya.deoscomputing.io/threat-model/v1 Maintainer: DEOS Computing (did:web:deos.computing)

The consolidated threat model for the v1.1 substrate. Extends and references the per-spec security-considerations sections, organized as a STRIDE-style matrix so a security reviewer can see the full picture in one place.

1. System under analysis

The Trust Substrate consists of:

• Receipt format and canonicalization — /receipts/format/v1
• MMR receipt log — /merkle/log/v1
• OTS anchoring — /anchor/v1, /anchor/bitcoin/v1
• Revocation log — /revocation/v1
• UCAN-shaped capability tokens + caveat algebra — /ucan/tokens/v1, /capability/caveats/v1
• did:deos identity method — /did/deos/v1
• Substrate signing-key rotation — /key-rotation/v1
• SCITT profile — /interop/scitt/v1
• MMR transparency — /mmr/transparency/v1
• Deployment topology, TLS termination, SigningProvider abstraction, FIPS migration — operator-internal specs not published publicly.

2. Trust boundaries

[ Issuer (agent, principal) ]
  │ signs Statements / UCAN tokens
  ↓
[ TLS proxy (operator-controlled) ]
  │ terminates TLS
  ↓
[ Substrate process ]                     ← attack surface 1: process compromise
  │ holds key.json or HSM-backed key      ← attack surface 2: secret exposure
  │ signs receipts + Receipts             ← attack surface 3: mis-signing
  │ MMR + persistence                     ← attack surface 4: log corruption
  ↓
[ OTS calendar(s) ]                       ← attack surface 5: calendar lying
  │
  ↓
[ Bitcoin chain ]                         ← attack surface 6: chain reorg / SPV trust

[ Verifier ]                              ← attack surface 7: verifier compromise / cache poisoning

3. STRIDE matrix

3.1 Spoofing

3.2 Tampering

3.3 Repudiation

3.4 Information disclosure

3.5 Denial of service

3.6 Elevation of privilege

4. Threat model deltas vs SCITT architecture v22 §10

The SCITT architecture's threat model covers the SCITT-shaped surface generically. DEOS adds:

• UCAN-shaped capability layer: §3.6 wildcard escalation, caveat-revocation path injection, replay of UCAN tokens against different substrates. Mitigations are DEOS-specific (Registration Policy rule 3 in /interop/scitt/v1 §6.3).
• OTS anchoring rather than direct Bitcoin submission: §3.2 OTS calendar lying is a DEOS-specific threat; SCITT itself is anchor-agnostic. Mitigation: multi-rail anchoring per /anchor/bitcoin/v1 §4.
• did:deos resolution: §3.1 DID spoofing requires a substrate-specific resolver hierarchy that SCITT does not specify. Mitigations in /did/deos/v1 §13.

5. Out-of-scope threats

The substrate explicitly does NOT defend against:

• Compromise of the issuer's signing key (the principal's key, not the substrate's). The substrate signs receipts under its own key; if the principal's key leaks, attacker can issue authentic-looking UCAN tokens. Mitigation is the principal's responsibility (their own key management); the substrate's revocation log lets them retire compromised tokens.
• Compromise of an OTS calendar's Bitcoin-submission key. The substrate trusts the calendar to faithfully submit. Mitigation: multi-rail anchoring + Bitcoin-side verification (v1.2).
• Bitcoin protocol-level compromise (51% attack, consensus-level reorg). The substrate's timestamping is only as good as Bitcoin; out of substrate scope.
• Side-channel attacks on the host hardware (Spectre, Meltdown, etc.). Operator's host hardening is out of substrate scope.
• Power-grid / supply-chain compromise of the underlying compute. Out of substrate scope.

6. Threat-actor profiles

6.1 Curious external observer

Goal: read receipts. Capabilities: HTTP requests to public endpoints. Defense: receipts are public by design; hashes-only payloads contain no sensitive content.

6.2 Network MITM

Goal: substitute a fake substrate's identity to verifiers. Capabilities: control over network path between verifier and substrate. Defense: TLS with cert pinning (operator TLS-termination spec).

6.3 Compromised operator

Goal: rewrite history, suppress evidence, mis-issue receipts. Capabilities: full control of substrate process + secret key. Defense: append-only MMR + OTS anchoring + log gossip + log-of-logs make rewriting detectable. HSM/KMS migration limits secret exposure. Cannot fully prevent — gossip + multi-witness is the structural answer.

6.4 Malicious issuer

Goal: forge actions claiming authority they don't have. Capabilities: any agent with a UCAN token. Defense: Registration Policy chain validation rejects mis-shaped chains; capability narrowing rules prevent privilege escalation.

6.5 Insider with substrate disk access

Goal: extract secrets, plant tampered state. Capabilities: file-system read on the substrate's host. Defense: file-permissions + filesystem encryption + HSM/KMS for secrets. Tampered state detected at startup via snapshot checksums.

6.6 Compromised verifier (downstream)

Goal: misrepresent verification results to a third party. Capabilities: verifier process compromise. Defense: verifier signs its own attestations of what it verified; downstream parties consume the verifier's signed attestations and apply their own pinning. Out of substrate scope.

7. Test coverage

External penetration testing covers the threat model. The team is asked to:

• Confirm each threat in §3 has a working mitigation.
• Identify any threats in §3 whose claimed mitigation does not actually hold.
• Identify any threats not in §3 that should be.

A conformance vector suite exercises the cryptographic primitives' edge cases. A fuzzing harness exercises canonicalization, MMR, UCAN parser, and caveat evaluation against random inputs.

8. References

• RFC 3552 — Internet security considerations
• STRIDE threat model framework (Microsoft)
• Each published substrate spec carries its own §-level security considerations that supplement this consolidated view.

9. Maintainer

DEOS Computing — threat-model updates and pen-test reports to github.com/DEOS-Computing.

License: CC BY 4.0 (text), Apache-2.0 (reference code).