Threat Analysis

​

Economic Attack Vectors

​

1. 51% Stake Attack

• High stake requirements make this expensive
• Slashing reduces profitability of attack
• Reputation damage is permanent
• Community can fork if attack detected

If total staked = 1M tokens
Attack requires >500K tokens
Plus: slashing risk, opportunity cost

​

2. Verifier Cartel

• Randomized committee selection
• MAD-based outlier rejection
• Multiple independent verifiers required
• On-chain audit trail enables detection

​

3. Score Manipulation for Bribes

• Anonymous commit-reveal hides scores
• Multiple verifiers must be bribed
• Outlier detection catches suspicious patterns
• Honest verifiers profit from accurate scores

​

Technical Attack Vectors

​

4. Front-Running Attacks

• Commit-reveal protocol
• Hash includes randomness
• Commits timestamped on-chain

​

5. Evidence Manipulation

• DataHash commits to evidence root
• Merkle proofs verify inclusion
• Immutable storage (Arweave)
• DKG signatures prevent tampering

​

6. Timestamp Manipulation

• VLC makes ancestry tampering detectable
• Tolerance limits on clock drift
• On-chain anchoring of timestamps
• Multiple verifiers check consistency

​

7. Sybil Identities

• ERC-8004 registration requires unique address
• Stake per identity raises cost
• New accounts have low reputation weight
• Pattern detection in scoring

​

Operational Attack Vectors

​

8. DoS on Verifiers

• Fee market for submissions
• Rate limiting per address
• Committee sampling limits workload
• Multiple verifier networks

​

9. Evidence Withholding

• Evidence availability requirements
• Multiple storage providers required
• Challenge-response mechanism
• Slashing for unavailable evidence

​

10. Griefing Attacks

• Submission fees
• Stake slashing for invalid submissions
• Quick pre-verification checks
• Reputation requirements for submission

​

Attack Scenarios

​

Scenario A: Malicious Worker

1. Alice submits fabricated work evidence
2. Creates fake DKG with inflated contributions
3. Verifiers audit and detect inconsistencies
4. Causal audit fails (missing signatures, broken VLC)
5. Work rejected, Alice loses stake

​

Scenario B: Verifier Collusion

1. Eve and Mallory collude to inflate Bob's scores
2. They coordinate to submit [95, 95, 95, 95, 95]
3. Honest verifiers submit [70, 75, 72, 80, 78]
4. MAD-based aggregation:
   - Median: ~75
   - MAD: ~3
   - Eve/Mallory's 95s are >3×MAD away
   - Their scores are rejected as outliers
5. Consensus reflects honest verifiers
6. Eve/Mallory are slashed for deviation

​

Scenario C: Bribery Attack

1. Worker offers verifiers $100 to score 90+
2. Commit-reveal means verifiers can't see others' commits
3. Some accept bribe, some don't
4. On reveal:
   - Bribed verifiers: [92, 94, 91]
   - Honest verifiers: [70, 72, 68, 71]
5. Bribed verifiers are outliers → slashed
6. Bribery cost + slashing loss > potential gain

​

Detection Mechanisms

​

Security Monitoring

​

On-Chain Metrics

• Score distribution per verifier
• Slashing frequency
• Committee composition over time
• Evidence availability rates

​

Off-Chain Monitoring

• Verifier behavior patterns
• Network communication analysis
• Storage provider availability
• Cross-reference with external data

​

Related