Chapter 11: Security and Anti-Cheating

11.1 Threat Model

Threat	Description	Impact
Merit farming	Obtaining MeriToken through fake contributions	Inflated voting power
Collusive voting	Multiple parties conspiring to manipulate recognition votes	Illegitimate Merit acquisition
Intimacy farming	Fabricating interactions to boost intimacy	Bypassing exclusions, reducing inheritance attenuation
Identity forgery	Creating fake HumanIDs	Multiple identities acquiring multiple Merit shares
Sybil attack	One person controlling multiple identities	Manipulating votes

11.2 Preventing Merit Farming

Safeguards for Objective Measurement

• System records automatically, leaving little room for human manipulation
• Cross-verification is possible (e.g., comparing work hours vs. output)
• Statistical anomaly detection

Safeguards for Subjective Evaluation

Core principle: make the cost of cheating far exceed the benefit.

1. Intimacy exclusion: exclude voters with close relationships
2. MeriToken weighting: high-reputation voters carry more weight; cheaters must first accumulate substantial genuine reputation
3. Behavioral auditing: frequently voting in favor of a specific individual → flagged as anomalous
4. Random sampling: randomly selecting voters to reduce the possibility of collusion
5. Retroactive accountability: once cheating is discovered, all participants are penalized

11.3 Preventing Intimacy Farming

• Interaction quality assessment (not just frequency)
• One-way interactions are invalid (must be bidirectional)
• Large volumes of interactions in a short period are treated as anomalous
• Isolated high-frequency interactions between two individuals (with no shared social circle) are treated as suspicious

11.4 Key Security

• Multi-signature schemes: critical operations require confirmation from multiple keys
• Key rotation: periodic replacement
• Social recovery: trusted contacts assist in recovery

11.5 Privacy Protection

• Voting content is not public (ZKP); only results are disclosed
• Intimacy values can be selectively disclosed
• Interaction content is not stored on-chain
• Anonymous participation is supported (ZKP proves eligibility without revealing identity)

11.6 Discussion Notes

Design philosophy of the security mechanism:

• There is no perfect anti-cheating solution; the goal is to make the cost of cheating far exceed the benefit
• Multi-layered defenses are more effective than any single mechanism
• Preventive measures + retroactive accountability form a closed loop
• Anti-cheating is a continuous adversarial process; the system must be able to evolve