Paper 2023/1059

Provably Secure Blockchain Protocols from Distributed Proof-of-Deep-Learning

Abstract

Proof-of-useful-work (PoUW), an alternative to the widely used proof-of-work (PoW), aims to re-purpose the network's computing power. Namely, users evaluate meaningful computational problems, e.g., solving optimization problems, instead of computing numerous hash function values as in PoW. A recent approach utilizes the training process of deep learning as ``useful work''. However, these works lack security analysis when deploying them with blockchain-based protocols, let alone the informal and over-complicated system design. This work proposes a distributed proof-of-deep-learning (D-PoDL) scheme concerning PoUW's requirements. With a novel hash-traininßg-hash structure and model-referencing mechanism, our scheme is the first deep learning-based PoUW scheme that enables achieving better accuracy distributively. Next, we introduce a transformation from the D-PoDL scheme to a generic D-PoDL blockchain protocol which can be instantiated with two chain selection rules, i.e., the longest-chain rule and the weight-based blockchain framework (LatinCrypt' 21). This work is the first to provide formal proofs for deep learning-involved blockchain protocols concerning the robust ledger properties, i.e., chain growth, chain quality, and common prefix. Finally, we implement the D-PoDL scheme to discuss the effectiveness of our design.