Proof-of-Learning with Incentive Security

2024-04-13

arXiv.org

AI Summary

Traditional Proof-of-Work (PoW) suffers from excessive energy consumption, while Proof-of-Useful-Work (PoUW) variants face security vulnerabilities and high verification overhead. Method: This paper proposes the first Proof-of-Learning (PoL) consensus framework with provably incentive-compatible security, natively embedding machine learning training tasks—e.g., SGD optimization—into the consensus protocol, synergistically integrating zero-knowledge verifiable computation, game-theoretically designed incentive-compatible mechanisms, and a lightweight blockchain protocol extension. Contributions: (1) Introduces the “incentive security” paradigm, ensuring protocol robustness without relying on Byzantine assumptions—even when both task publishers and verifiers are untrusted; (2) Breaks the theoretical lower bound on verification cost, achieving O(log E / E); (3) Provides provable resistance against double-spending and double-voting attacks, enabling a fully decentralized AI compute marketplace; (4) Delivers the first practical, secure, and economically rational PoL solution for green blockchain systems.

Technology Category

• Machine Learning: Efficient ML / Green AI
• Search and Optimization: Learning to Search
• Multiagent Systems: Mechanism Design

Application Category

• Economics, Online Markets and Human Computation: Economic aspects of blockchain and cryptocurrencies
• Security and Privacy: Blockchains and distributed ledgers
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Most concurrent blockchain systems rely heavily on the Proof-of-Work (PoW) or Proof-of-Stake (PoS) mechanisms for decentralized consensus and security assurance. However, the substantial energy expenditure stemming from computationally intensive yet meaningless tasks has raised considerable concerns surrounding traditional PoW approaches, The PoS mechanism, while free of energy consumption, is subject to security and economic issues. Addressing these issues, the paradigm of Proof-of-Useful-Work (PoUW) seeks to employ challenges of practical significance as PoW, thereby imbuing energy consumption with tangible value. While previous efforts in Proof of Learning (PoL) explored the utilization of deep learning model training SGD tasks as PoUW challenges, recent research has revealed its vulnerabilities to adversarial attacks and the theoretical hardness in crafting a byzantine-secure PoL mechanism. In this paper, we introduce the concept of incentive-security that incentivizes rational provers to behave honestly for their best interest, bypassing the existing hardness to design a PoL mechanism with computational efficiency, a provable incentive-security guarantee and controllable difficulty. Particularly, our work is secure against two attacks, and also improves the computational overhead from $Theta(1)$ to $O(frac{log E}{E})$. Furthermore, while most recent research assumes trusted problem providers and verifiers, our design also guarantees frontend incentive-security even when problem providers are untrusted, and verifier incentive-security that bypasses the Verifier's Dilemma. By incorporating ML training into blockchain consensus mechanisms with provable guarantees, our research not only proposes an eco-friendly solution to blockchain systems, but also provides a proposal for a completely decentralized computing power market in the new AI age.

Problem

• Blockchain
• Useful Work Proof (PoUW)
• Machine Learning

Innovation

• Incentive Security
• Machine Learning Proof of Work
• Blockchain Environmental Sustainability