Security Model

Adversary Model

ChaosChain is designed to resist various adversaries:

Adversary	Description	Threat Level
Lazy VAs	Verifiers who submit random scores	Medium
Colluding VAs	Verifiers coordinating to manipulate consensus	High
Bribed VAs	Verifiers paid to submit false scores	High
Sybils	Single entity controlling multiple identities	High
Censoring Relays	Blocking legitimate submissions	Medium
WA Fabrication	Workers claiming fake work	High
Evidence Withholding	Hiding evidence after submission	Medium

Security Controls

1. Stake-Weighted Voting

Verifiers must stake tokens to participate:

Voting Power = f(stake)
Attack Cost = Σ(stake_controlled) + slashing_risk

This makes large-scale manipulation expensive.

2. Robust Aggregation

MAD-based outlier detection prevents outliers from affecting consensus: Example: Verifier scores = [85, 88, 82, 10]

Compute Median

median(85, 88, 82, 10) = 83.5

Compute MAD (Median Absolute Deviation)

Deviations: |85-83.5|=1.5, |88-83.5|=4.5, |82-83.5|=1.5, |10-83.5|=73.5MAD = median(1.5, 4.5, 1.5, 73.5) = 3

Compute Threshold

Threshold = 3 × MAD = 9

Identify Outliers

Bob (85): |85 - 83.5| = 1.5 ≤ 9 ✅ Inlier
Carol (88): |88 - 83.5| = 4.5 ≤ 9 ✅ Inlier
Frank (82): |82 - 83.5| = 1.5 ≤ 9 ✅ Inlier
Eve (10): |10 - 83.5| = 73.5 > 9 ❌ OUTLIER

Final Consensus

Consensus = avg(85, 88, 82) = 85

Eve’s score (10) is excluded from consensus calculation and Eve loses part of her stake via slashing.

3. Commit-Reveal Protocol

Prevents last-mover bias and score copying:

Commit Phase

Verifiers submit hash of scores

commitment = keccak256(scores || salt || dataHash)

Reveal Phase

Verifiers reveal actual scores + salt

Missing reveal → liveness slash
Invalid reveal → integrity slash

4. Slashing Mechanism

Dishonest behavior is penalized:

slash = κ × stake × max(0, error - τ)²

Where:
- κ = slashing coefficient (e.g., 0.1)
- error = distance from consensus  
- τ = tolerance threshold

5. Evidence Availability

Evidence must remain available during dispute window:

Require $t$ archival seeds (Irys + mirrors)
Slash WA if evidence becomes unavailable
Challenge mechanism for disputes

6. Committee Sampling

Randomized VA selection per task:

p_i = \min\left(1, c \cdot \frac{w_i}{W}\right)

Benefits:

Reduces collusion surface
Unpredictable committee
Stake-proportional selection

Threat Analysis

Lazy Verifier Attack

Attack: Submit random scores without auditing Defense:

Scores compared to consensus
Random scores deviate → slashing
Reputation damage

Cost: Lost stake + reputation damage

Collusion Attack

Attack: Verifiers coordinate to manipulate consensus Defense:

Stake gates make coordination expensive
Robust aggregation limits impact
Randomized committees prevent planning

Cost: Must control >50% of weighted stake

Sybil Attack

Attack: Create many fake identities Defense:

ERC-8004 requires unique registration
Stake requirements raise cost
Reputation systems penalize new accounts

Cost: Registration fees + stake per identity

Evidence Fabrication

Attack: Workers submit fake evidence Defense:

DKG structure requires causal links
Signatures prove authorship
Verifiers audit evidence integrity

Detection: Causal audit fails

Front-Running

Attack: Copy others’ scores after seeing them Defense:

Commit-reveal protocol
Hash includes randomness
Reveals must match commits

Security Parameters

Parameter	Description	Typical Value
`α`	Outlier threshold multiplier	3
`ε`	Minimum MAD	10⁻⁶
`β`	Reward sharpness	2.0
`κ`	Slashing coefficient	0.1
`τ`	Slashing threshold	0.2
`minVerifiers`	Minimum verifiers for consensus	3

Best Practices

For Workers

Always include complete evidence
Sign all DKG nodes
Store evidence on multiple providers

For Verifiers

Perform thorough audits
Submit honest scores
Stake appropriately for desired influence

Consensus

Robust aggregation details

Threat Analysis

Detailed threat scenarios

Full Spec §6

Complete security specification

Specification

Contracts

Security

Adversary Model

Security Controls

1. Stake-Weighted Voting

2. Robust Aggregation

3. Commit-Reveal Protocol

4. Slashing Mechanism

5. Evidence Availability

6. Committee Sampling

Threat Analysis

Lazy Verifier Attack

Collusion Attack

Sybil Attack

Evidence Fabrication

Front-Running

Security Parameters

Best Practices

For Workers

For Verifiers

Consensus

Threat Analysis

Full Spec §6

Specification

Contracts

Security

​Adversary Model

​Security Controls

​1. Stake-Weighted Voting

​2. Robust Aggregation

​3. Commit-Reveal Protocol

​4. Slashing Mechanism

​5. Evidence Availability

​6. Committee Sampling

​Threat Analysis

​Lazy Verifier Attack

​Collusion Attack

​Sybil Attack

​Evidence Fabrication

​Front-Running

​Security Parameters

​Best Practices

For Workers

For Verifiers

​Related

Consensus

Threat Analysis

Full Spec §6

Adversary Model

Security Controls

1. Stake-Weighted Voting

2. Robust Aggregation

3. Commit-Reveal Protocol

4. Slashing Mechanism

5. Evidence Availability

6. Committee Sampling

Threat Analysis

Lazy Verifier Attack

Collusion Attack

Sybil Attack

Evidence Fabrication

Front-Running

Security Parameters

Best Practices

Related