Skip to main content

Threat Model & Adversarial Analysis

1. Purpose and Scope

This annex provides a security-oriented analysis of the VeriSeal proof infrastructure under adversarial conditions.

It defines:

  • Threat actors and capabilities
  • Trust boundaries
  • Attack surfaces
  • Security assumptions
  • Detection and mitigation mechanisms
  • Residual risks and governance requirements

This annex is written for institutional stakeholders (CISO, Risk, Audit, Compliance).


2. System Assets to Protect

VeriSeal is designed to protect integrity and verifiability of evidence artifacts and proof records.

Primary assets:

  1. Artifact Integrity The sealed artifact must remain verifiable and tamper-evident over time.

  2. Proof Object Integrity Proof records (canonical JSON) must not be alterable without detection.

  3. Ledger Continuity Append-only chain must remain structurally immutable and chronologically consistent.

  4. Verification Independence Third parties must be able to verify proofs without operator trust.

  5. Timestamp Integrity (optional external anchoring) When external anchoring exists, the anchor must strengthen time certainty.

  6. Key Integrity (signing keys) Private keys used for signing must remain confidential and governed.


3. Threat Actors

3.1 External Adversary

Capabilities:

  • Network-level interception attempts
  • Public endpoint probing
  • Artifact substitution attempts
  • Replay attempts for capture flows (when applicable)

Constraints:

  • No direct access to signing keys or internal storage (assumed)

3.2 Insider Threat (Operator or Privileged Admin)

Capabilities:

  • Potential access to infrastructure
  • Attempts to alter records
  • Attempts to delete or selectively hide evidence
  • Attempts to manipulate PDF or presentation outputs

3.3 Malicious Evidence Submitter

Capabilities:

  • Submit false or misleading content
  • Submit manipulated artifacts
  • Attempt to exploit workflow ambiguity
  • Attempt repudiation after sealing ("I never sent this")

3.4 Compromised Client Device

Capabilities:

  • Malware modifies artifact before sealing
  • Credential compromise
  • Intercepts or alters local data before upload

3.5 Supply Chain / Dependency Adversary

Capabilities:

  • Dependency compromise
  • Build pipeline compromise
  • Runtime injection attempts

4. Trust Boundaries

VeriSeal enforces explicit trust boundaries:

Inside VeriSeal trust boundary:

  • Hash computation
  • Canonical proof object generation
  • Ledger append operation
  • Signature generation (controlled key domain)
  • Optional anchor submission

Outside trust boundary:

  • Artifact creation tools
  • User device security
  • Semantic truth of content
  • Legal interpretation
  • Human intent and context

Key principle: VeriSeal secures structure, not meaning.


5. Attack Surface Overview

Primary attack surfaces:

  1. Input artifact ingestion (upload / payload generation)
  2. Proof generation pipeline (hashing + canonicalization + signing)
  3. Ledger storage (append-only chain integrity)
  4. Verification endpoints (public JSON, PDF render, HTML verify page)
  5. Key management (private key confidentiality and usage policy)
  6. Build/CI and dependency chain (supply chain integrity)

6. Threat Scenarios and Mitigations

6.1 Post-Sealing Artifact Tampering

Attack: Artifact modified after sealing.

Detection:

  • Recomputed SHA-256 digest differs.

Mitigation:

  • Deterministic hashing
  • Proof validation requires recomputation

Residual risk:

  • None for detection (integrity always detects change)
  • Semantic disputes remain possible

6.2 Proof Object Tampering (JSON)

Attack: Alter proof JSON fields (timestamps, identifiers, flags).

Detection:

  • Signature verification fails
  • Hash mismatch in canonical recomputation

Mitigation:

  • Canonical JSON as source-of-truth
  • Signature over canonical payload

Residual risk:

  • If signing key is compromised (see key compromise scenario)

6.3 Ledger Mutation / Deletion

Attack: Delete an entry or rewrite historical chain.

Detection:

  • Chain discontinuity (previous_entry_hash mismatch)
  • entry_hash recomputation mismatch

Mitigation:

  • Append-only architecture
  • Chain hash dependency
  • Regular external anchoring recommended for high assurance

Residual risk:

  • Storage-level attacks can delete entire ledger if governance is weak (availability risk; integrity detection remains, but evidence may be missing)

6.4 Selective Evidence Suppression ("Evidence Hiding")

Attack: Operator hides or refuses to serve a proof endpoint.

Detection:

  • Public reference exists but endpoint not responding
  • External copies of JSON prove prior existence

Mitigation:

  • Encourage institutional archiving of proof bundles
  • External anchoring strengthens existence claims
  • Governance policies for availability

Residual risk:

  • Availability remains operationally governed

6.5 Presentation Layer Manipulation (PDF/HTML)

Attack: Modify PDF rendering or HTML view to show false status.

Detection:

  • Canonical JSON remains source-of-truth
  • Verifier recomputes hashes and validates signature

Mitigation:

  • Render-only rule
  • Verification logic never depends on presentation output

Residual risk:

  • Social engineering risk remains (users trusting visuals without verifying)

6.6 Replay Attacks (Capture-Based Proofs)

Attack: Replay a prior video/audio capture to fake liveness.

Detection/Mitigation (when capture flows exist):

  • Challenge-response binding (dynamic prompts)
  • Runtime event log hashing
  • Media binding to runtime payload

Residual risk:

  • Dependent on capture workflow quality
  • Not applicable to static artifact sealing use cases

6.7 Man-in-the-Middle / Transport Layer Attack

Attack: Intercept or alter traffic during upload or verification.

Mitigation:

  • TLS enforced (HTTPS)
  • Hash recomputation and signature validation prevents silent tampering

Residual risk:

  • Credential theft is outside cryptographic model
  • Network downgrade attacks mitigated via strict HTTPS and HSTS (recommended)

6.8 Private Key Compromise

Attack: Signing key stolen.

Impact:

  • Attacker can issue forged proofs that appear valid
  • Historical proofs remain verifiable but trust in signer is impacted

Mitigation:

  • HSM / KMS-backed signing
  • Key rotation policy
  • Audit logs for signing operations
  • Separation of duties (dual control)

Residual risk:

  • Highest-impact scenario; governance is mandatory

6.9 Hash Function Weakening (Cryptographic Breakthrough)

Attack: Practical collision or second preimage against SHA-256.

Impact:

  • Integrity assumptions degrade

Mitigation:

  • Algorithm agility (hash_version field)
  • Ability to re-anchor proofs using stronger primitives
  • Support for dual-hash strategies in future versions

Residual risk:

  • Long-term cryptographic evolution risk exists for all systems

6.10 Supply Chain Compromise

Attack: Build dependencies inject malicious logic.

Mitigation:

  • Pin dependencies
  • Reproducible builds
  • Code signing
  • SBOM generation
  • CI integrity controls

Residual risk:

  • Requires disciplined software governance

7. Security Assumptions

VeriSeal security assumptions:

  • SHA-256 remains secure (no feasible collisions)
  • Signature scheme remains secure (no feasible forgery without private key)
  • Canonical serialization is deterministic and controlled
  • Append-only enforcement is operationally protected
  • Private keys are protected by institutional key management

8. Security Guarantees (What VeriSeal Actually Guarantees)

VeriSeal guarantees:

  • Detection of any post-sealing artifact modification
  • Deterministic verification reproducibility
  • Structural integrity and chain continuity (if ledger available)
  • Proof object tamper-evidence via signature
  • Optional external time anchoring reinforcement

VeriSeal does not guarantee:

  • Truth of content
  • Identity legitimacy by default
  • Protection against compromised client devices
  • Availability of proof endpoints without governance

9. Residual Risk and Governance Requirements

VeriSeal is strongest when governance enforces:

  • Key management policy (HSM/KMS, rotation, audit)
  • Availability policy (archival bundles, redundancy)
  • Role segregation (sealing vs verification vs administration)
  • Incident response procedures
  • External anchoring policy for high assurance use cases

Without governance, integrity remains detectable, but operational resilience may degrade.


10. Institutional Security Posture Summary

VeriSeal is designed as:

  • A deterministic integrity protocol
  • An audit-compatible evidence layer
  • A system minimizing trust dependency
  • A tamper-evident ledger + proof object model

It provides verifiable integrity guarantees under adversarial conditions, provided that key governance and operational controls are enforced.


11. Risk Classification Matrix

This section formalizes threat severity according to impact and likelihood.

Risk classification model:

  • Likelihood: Low / Medium / High
  • Impact: Moderate / Significant / Critical
  • Residual Risk: After mitigation controls
  • Risk Owner: Governance responsibility domain
Threat ScenarioLikelihoodImpactMitigation StrengthResidual RiskRisk Owner
Artifact tampering post-sealMediumSignificantCryptographic detectionLowVerification layer
Proof JSON alterationLowSignificantSignature + canonicalizationLowCryptographic layer
Ledger mutationLowCriticalHash chaining + detectionLowInfrastructure governance
Evidence suppressionMediumSignificantArchival + anchoringMediumOperational governance
Presentation manipulationMediumModerateJSON source-of-truthLowVerification discipline
Private key compromiseLowCriticalHSM/KMS + rotationMediumKey governance
Hash algorithm weakeningVery LowCriticalAlgorithm agilityLowCryptographic governance
Supply chain compromiseMediumSignificantCI controls + SBOMMediumDevSecOps governance

Risk posture summary:

  • Structural integrity risks are strongly mitigated.
  • Key compromise remains highest-impact scenario.
  • Governance maturity directly influences residual risk.

12. Governance & Role Segregation Model

VeriSeal security assumes institutional separation of duties.

12.1 Functional Roles

Sealing Operator

  • Initiates proof generation.
  • Cannot modify historical ledger entries.

Verification Authority

  • Performs independent validation.
  • Has no access to signing keys.

Key Custodian

  • Manages signing keys.
  • Operates under dual control policy.

Infrastructure Administrator

  • Manages system availability.
  • Cannot alter signed proof payloads.

Audit Function

  • Reviews logs and chain continuity.
  • Independent from operations.

12.2 RACI Overview

FunctionSealVerifySignMaintainAudit
Sealing OperatorRC---
Verification Authority-R--C
Key Custodian--R-C
Infrastructure Admin---RC
Audit-CCCR

R = Responsible C = Control / Oversight


12.3 Governance Principle

Integrity is enforced by cryptography. Trust is enforced by governance separation.

Cryptographic guarantees reduce reliance on human trust, but institutional governance ensures resilience.


13. Incident Response & Key Compromise Procedure

This section defines minimum institutional response posture.

13.1 Key Compromise Scenario

If signing key compromise is suspected:

  1. Immediate key revocation.
  2. Stop new proof issuance.
  3. Generate new signing key pair.
  4. Publish public incident notice.
  5. Re-anchor system state (if applicable).
  6. Conduct forensic audit.

Historical proofs remain structurally verifiable. Trust domain shifts to key validity timeline.


13.2 Proof Integrity Incident

If ledger corruption is detected:

  • Recompute chain continuity.
  • Identify divergence point.
  • Restore from validated backup.
  • Publish integrity statement.

13.3 Public Transparency Principle

For institutional deployments:

  • Incident disclosure policy required.
  • Timestamped incident logs recommended.
  • Independent audit advisable for high assurance domains.

13.4 Long-Term Cryptographic Evolution

If hash or signature primitives weaken:

  • Introduce versioned hash field.
  • Re-anchor existing proofs.
  • Support dual-hash strategy during migration.

Algorithm agility must be built into governance roadmap.