Research Of A New Consensus Mechanism Based On The Computing Collaboration Of Artificial Intelligence And Blockchain

Proof of Work(PoW)consensus mechanism is widely used in various blockchain systems.In a PoW-based blockchain system,the miners have to devote a large amount of computing resource to a brute-force search for a random number(Nonce),and a valid nonce should make the hash value of the block header smaller than the target hash value.However,PoW is disputed for tremendous computing resource waste,because the intermediate results obtained by performing a large number of hash calculations in the consensus process don't have application value.Meanwhile,the artificial intelligence(AI)training,which can be represented by machine learning,has massive computing demands but often faces a lack of computing power.If we can find a link between blockchain and AI to realize computing power coordination between these two systems,then this dilemma will be solved.However,the current studies on the computing power problem of blockchain and AI is still insufficient,and the amount of the salvaged computing power is limited.In order to solve this problem,this thesis first summarizes the current research status of the computing power problems in PoW-based blockchain and AI,then analyzes the main problems faced by current studies.Based on these analyses,this thesis proposes a system framework and specific technologies for the deep integration and efficient reuse of the computing power in PoW-based blockchain and AI.The availability and efficiency of the proposed framework and technologies are verified by experiments.The main research contents are as follows:(1)Taking matrix multiplication calculation(MMC)as the link,this thesis combines the computing power demand of AI with the computing power supply of blockchain,and proposes a basic framework of computing collaboration between AI and blockchain(CCBAB).The core idea of this framework is that the blockchain's miners need to complete the MMC from AI's model training processes,so as to realize the computing collaboration between AI and blockchain.(2)Based on the basic framework of CCBAB,this thesis designs a new consensus algorithm,namely static-allocation based Evolved Proof-of-Work(s E-PoW),for the scenario that the computing power demand of AI and the blockchain miners are relatively stable.The related supporting algorithms,including the continuous multiplication processing mechanism,the result merging mechanism,the result verification mechanism and the malicious behavior punishment mechanism,are also proposed for the framework's operation.Based on s E-PoW consensus and the related supporting algorithms,the miners will perform MMC and hash calculation according to a static computing power allocation ratio during the consensus process,and the effective utilization of blockchain's computing power can be improved.(3)As for the scenario that AI's computing power demand and the blockchain miners are dynamic,this thesis designs another consensus algorithm,namely dynamic-allocation based Evolved Proof-of-Work(d E-PoW).In a d E-PoW based blockchain,the miners can flexibly adjust the proportion of their computing resources invested in MMC and hash calculation,and the system can guide the miners' behaviours when the amount of the external computing demand changes.While ensuring the computing efficiency of the miners,d E-PoW gives miners the flexibility to allocate computing resources and avoids resources mismatch.