Researchers have developed two Solidity smart contracts for federated learning reward systems, aiming to incentivize participation while maintaining data privacy. One contract distributes rewards based on a public quality score, while a privacy-preserving variant uses zero-knowledge proofs to hide individual scores. Experiments with the MNIST database revealed a trade-off: enhanced privacy led to decreased system robustness and increased vulnerability to adversarial attacks like Sybil attacks, ultimately resulting in honest contributors being excluded. AI
IMPACT Highlights a critical challenge in securing decentralized AI systems, potentially impacting the design of future federated learning reward mechanisms.
RANK_REASON Academic paper detailing a novel approach to federated learning reward systems and its empirical evaluation. [lever_c_demoted from research: ic=1 ai=1.0]
- EIP-7825
- federated learning
- Fusaka Oikawa
- MNIST database
- Solidity
- Sybil attacks
- zero-knowledge proofs
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
