Skip to content

Latest commit

 

History

History
27 lines (23 loc) · 4.7 KB

File metadata and controls

27 lines (23 loc) · 4.7 KB

TTP Threat Model

TTP expresses trust context. TTP alone does not enforce execution. Enforcement happens through RAP, Execution Exchange, FrontDesk-integrated gateways, API gateways, CI gates, or equivalent runtime controls.

Threat Description Impact Mitigation MVP Status Future Hardening
Forged trust claims Attacker creates a fake trust claim or edits score/timestamps. Unauthorized action may appear trusted. Validate issuers, signatures, schema, and immutable evidence references. Cleartext-dev only; not production safe. Signed claims, issuer registry, canonical signing, key rotation.
Stale trust proofs Old proof is reused after trust context has changed. Actions proceed on outdated trust. Expiration, freshness windows, short TTLs. Expiration and freshness semantics documented; basic checks implemented. Nonces, replay caches, runtime receipts.
Replay attacks Previously valid proof or receipt is replayed. Unauthorized repeated execution. Bind proof to action, resource, time, and nonce. Out of MVP. Nonce registry, receipt hash binding, RAP request binding.
Issuer compromise Valid issuer signs false claims. Trust model can be corrupted. Least privilege issuers, issuer reputation, revocation, independent issuers. Documented risk. Issuer registry, quorum rules, revocation feeds.
Malicious agent self-attestation Agent issues trust about itself. Inflated trust. Disallow self-attestation unless explicitly marked and low weight. Not enforced beyond examples. Issuer policy constraints and signed issuer metadata.
Trust farming Actor performs low-risk behavior to build trust for high-risk actions. Inappropriate trust transfer. Scope trust by action/domain/resource; use high-risk thresholds. Scope syntax exists. Risk-weighted decay, action-specific scoring.
Sybil attacks Many fake subjects or issuers inflate trust. Trust graph manipulation. Issuer validation, subject registration, governance controls. Out of MVP. Identity binding, issuer reputation, abuse detection.
Proof expiration bypass Runtime ignores expiration or clock is wrong. Expired trust accepted. Fail closed, reliable clocks, skew limits. Expiration checks implemented in reference CLI. Secure time, signed timestamps, RAP enforcement tests.
Policy downgrade attacks Lower proof mode or lower threshold is substituted. Weaker evaluation than intended. Pin proof requirements in authority context and runtime policy. Basic proof references parsed. Policy signatures and version pinning.
Cross-domain trust abuse Trust from one domain is reused in another. Scope escape. Domain and scope checks. Domain captured; limited enforcement. Domain-specific issuer policies and RAP mapping.
Runtime gate bypass Actor calls target system without RAP or gateway enforcement. TTP evaluation is skipped. Place enforcement at mandatory control points. Out of TTP core. Execution Exchange, service mesh, CI and API gateway adapters.
Receipt tampering Evidence or receipt is modified after creation. Audit and proof integrity loss. Sign receipts and hash evidence references. Output supports optional receipt hash. Canonical receipt schema and transparency log support.
Clock manipulation Actor or runtime shifts time to avoid decay or expiration. Trust remains valid too long or decays incorrectly. Trusted server-side evaluation time and skew limits. CLI uses evaluator time input. Secure time source and signed evaluation receipts.
Misconfigured trust thresholds Thresholds are too low or too broad. Weak authority control. Defaults, linting, review, risk-tier guidance. Examples use explicit thresholds. Policy linter and conformance tests.
Overbroad delegation Delegation transfers too much authority or lasts too long. Privilege expansion. Bound delegation by scope, max score, issuer, expiration. Delegation syntax example only. Delegation evaluator and cycle detection.
Trust issuer impersonation Attacker claims to be a trusted issuer. Fake claims accepted. Issuer identity verification and signature checks. Out of MVP proof mode. Issuer registry, JWKS/DID/PKI integrations.
Dependency compromise Parser or runtime dependency is compromised. Supply-chain compromise. Minimal dependencies, lockfiles, CI checks. CLI uses only Node standard library. SLSA provenance, signed releases, dependency scanning.

Security Boundary

TTP is a protocol and language for trust expression. It must be integrated with runtime enforcement to have security effect. If a system can bypass the RAP or gateway decision point, TTP cannot prevent execution.