Research On Blockchain Federated Learning Algorithm Based On DPoS

At present,the development of artificial intelligence technology is extremely rapid,in quite a number of application scenarios,artificial intelligence technology has been applied,but in the algorithm,computing power and other aspects are greatly improved,people also pay more attention to the protection of data privacy.Today’s machine learning algorithms increasingly rely on a large amount of data,and the development of ARTIFICIAL intelligence faces great challenges in privacy protection and data dispersion:data security,privacy protection and data islands.So Google first proposed the concept of federated learning in 2016.Federated learning is a deep learning algorithm that can train artificial intelligence without disclosing user data privacy.Federated learning has made outstanding achievements in protecting data privacy and solving the problem of data islands.However,regular federated learning is a centralized architecture,which faces many problems in real scenarios.For example,the security and robustness of regular federated learning under malicious node attack,the pressure of network communication,the utilization of network bandwidth,the influence of malicious data on the final efficiency of federated learning,and so on.Nowadays,in many cases,privacy leakage and data leakage are encountered in the process of communication transmission.First of all,data leakage greatly restricts the availability of big data.Although there is a large amount of data every day in the Internet era,there is still a lack of high latitude,high quality and data security.Secondly,every country is strengthening the protection of data security and privacy,so the protection of user data privacy will be the development trend of the world and the hot topic up to now.In view of the privacy security problems in data transmission,this paper proposes a study on blockchain federated learning algorithm based on DPoS.Will block chain technology combined with a federal study in the chain of blocks in consensus algorithm,and the realization of the DPoS algorithm firstly in this paper,based on the PoS has carried on the contrast experiment,the contrast between its two nodes trust and income situation,the proportion of malicious nodes as well as the proportion of honest nodes and so on,the experimental results show that PoS is obviously better than the DPoS algorithm under the condition of more than several algorithms.Secondly,the design and implementation of decentralized model verification,the design of reward mechanism and the selection of miners in blockchain federated learning are discussed.The innovation of this paper is based on DPoS blockchain federated learning,and an FL system supported by the credibility mechanism of BC platform is proposed.The design concept of the system is to introduce a voting-based mechanism to evaluate the reputation of each node.The reputation and reward are positively correlated,and any rational node will follow the model training rules to maximize its own interests.Theoretical analysis and simulation experiments show that the scheme can effectively solve the problems of privacy protection and data security in data transmission.This paper mainly studies the privacy disclosure in the process of data transmission and how to effectively protect it.Using blockchain as the infrastructure of federated learning can improve the security in the process of data transmission more reasonably and effectively.