Alleviating cybersecurity risks associated with centralized AI training and implementation is a burgeoning challenge in the construction industry. This paper addresses two primary questions: (1) What is the knowledge of AI security vulnerability in construction, and (2) How can AI be decentralized using blockchain? To this end, this paper proposes a blockchain-AI integrated framework (BAII), enabling AI to be trained, verified, and applied on a decentralized blockchain. The framework has been successfully validated in an excavator pose recognition scenario, demonstrating acceptable latency and high performance with 95 % accuracy, 94 % precision, and 96 % recall. This research is pivotal for construction managers and IT security professionals, enhancing the reliability and safety of AI applications in construction. The decentralized AI (DAI) approach can also inspire further research into motivating constructors to contribute to AI modeling and training through incentive mechanisms in the blockchain.
原文链接
分享
引用
