Research On Transaction Propagation Algorithms And Decentralized Machine Learning Framework For Blockchains

As a revolutionary technology,blockchain solves the problem of consensus under the participation of multiple parties and brings new choice for various industries.However,the transaction throughput limits the development of blockchain,and the transaction propagation speed of existing blockchain is much slower than that of block propagation,which threatens the security of blockchain.Machine learning has performed well in many applications,but it is constrained by a centralized third party updating the model,which hinders joint training of the model and open source sharing.Machine learning is expected to replace hashing algorithm as a useful proof-of-work in the future because of its unpredictable and verifiable characteristics.However,as the proof-of-useful-work,machine learning has the problem of long verification time,which affects the transmission speed of transactions or blocks.According to the above questions,the research work and contributions of this paper are as follows:First,we measure the transaction data of 366 blocks in the Bitcoin network,and the analysis shows that the verification time of transactions is the main reason that affects the efficiency of transaction propagation.Based on the characteristics of high aggregation coefficient and high average of the Bitcoin network,we propose a blockchain network structure called local clique network.In the local clique network,there are two kinds of relationships between nodes and neighbors,i.e.,insiders and outsiders,where for insider and outsider group we use two kinds of strategies to forward transactions.Experimental results show that the local clique network can effectively reduce the message redundancy,significantly improve the transmission efficiency of transaction,and has good robustness under the condition of high node average degree.Secondly,this paper proposes and implements the data and model chain(DMChain),by utilizing computational power consumption in training neural network with proof-of-work.As a blockchain that can be used to share data and machine learning models,the data shared anonymously by the whole network node are used in model chain,and the neural network model is explored based on the primary network,thus realizing neural network model update without relying on any third-party.The shared data is signed with a ring signature to protect local data privacy.The whole network uses the same test set to evaluate the model,and the adopted model can be regarded as proof-of-work.The paper proposes two reward mechanisms,i.e.,material reward and model reward.To deal with potential threats,e.g.,blockchain ledger analysis,dirty data attacks and fraudulent voting,this paper proposes ideal ring signature scheme and several solutions.Extensive experiments on real data are conducted,and the results show that the model in the model chain can adapt to the user changes and data changes.Finally,we apply the local clique network to the DMChain and improve the propagation of model transactions in the DMChain.Propagation experiments show that the local clique network can effectively reduce propagation delay due to validation,which is of practical significance for promoting the use of machine learning costs as proof-of-useful-work.