Research On Abnormal Transaction Identification And Traceability Method For Blockchain

With the rapid rise of blockchain currency system and its own decentralized,anonymous and other characteristics,blockchain has become one of the tools for criminals to evade sanctions.As a result,money laundering,Ponzi schemes,pyramid schemes and other unusual transactions have sprung up on blockchain.For blockchain technology,abnormal transactions can be effectively identified and traced back to the source,so that regulators can effectively control and manage the blockchain digital currency system and other applications,which is the key to the development and promotion of blockchain technology.However,there are the following problems in the process of identifying and tracing abnormal transactions in blockchain:1.The imbalance between normal and abnormal transaction data will lead to poor results in the training process of neural network; The problem of unknown transaction characteristics is that the characteristics with strong expression ability of the model can be mined from the transaction characteristics to improve the ability of the model to identify abnormal transactions.2.The convergence speed of the central algorithm is slow.The traditional central exception recognition algorithm requires high-performance computing equipment,which affects the efficiency of blockchain abnormal transaction recognition.In addition,the central algorithm may be accompanied by the risk of user privacy disclosure,which weakens the enthusiasm of users to participate in model training.3.Currently,most researches on blockchain transaction supervision technology focus on transaction data analysis.The field of blockchain abnormal transaction traceability is in its initial stage and there is a lack of relevant theoretical research.From two perspectives of blockchain abnormal transaction identification and tracing,this thesis proposes feasible methods to deal with the above challenges.The main contributions and innovations of this thesis are as follows:1.Based on feature engineering and self-service tags,a method was designed to improve the recognition of blockchain abnormal transactions,fully extract the expression ability of key features of blockchain transactions,increase the proportion of abnormal transactions based on feature similarity,make full use of unknown transaction attribute data,reduce the impact caused by sample imbalance,and increase the recall rate of 43%and F1＿score of 22%.2.A block chain abnormal transaction recognition model based on federated learning is designed.The edge nodes in the block chain network are trained collaboratively on the abnormal transaction recognition model by using federated learning,which protects user privacy and increases the convergence speed of the model by 1.5 times.3.The propagation law of block chain abnormal transaction in the network is studied,and the mathematical propagation model of block chain abnormal transaction is constructed.A block chain abnormal transaction tracing method based on the graph neural network is designed.The propagation probability between nodes is calculated by using the graph neural network,and the noise unrelated to the abnormal transaction source node is eliminated.Compared with the comparison method,F1＿score is improved by at least 12%.