Research And Design Of Bitcoin Abnormal Behavior Detection System

Abnormal behavior refers to behaviors that have significant deviations in patterns and characteristics compared with most observations.In the Bitcoin blockchain:abnormal fluctuations in traffic have significant numerical deviations,and the behavior patterns of illegal behaviors are different.They both represent a typical abnormality,and such abnormal behaviors often imply danger signals that threaten the system or cause huge market fluctuations.Therefore,the application of abnormal behavior detection in the Bitcoin blockchain is conducive to early perception of risks and response,effectively combating illegal and criminal activities,and preventing them from harming the stability and integrity of the network.But at present,there is no obvious superiority method for Bitcoin application scenarios.Based on the public data generated by the Bitcoin network,this thesis studies and designs a set of abnormal behavior detection systems.Starting from the macro traffic data and micro specific transactions,taking into account the overall network and specific users,two different anomaly detection methods are designed:(1)Taking macroscopic network traffic behavior as the research object: use a combination of support vector machines and codecs to provide visual analysis results and abnormal alarms for unsupervised network macroscopic traffic data.The encoder and decoder are responsible for the reconstruction and visualization of information.One class support vector machine determines abnormalities based on the deviation between the real value and the reconstructed value.Finally,the information reconstruction ability,visualization effect,and response to major events of the scheme are tested and compared.(2)Taking micro-transaction behavior as the research object: aiming at illegal types of abnormal transactions,based on transaction characteristics and transaction graph information,it is proposed to use the evolving graph convolutional neural network and graph attention mechanism to extract features.Random forest is used for abnormal behavior detection and judgment to provide warning.Finally,the classification ability of the model was tested.In the final integrated system,the reconstruction results of the codec can more intuitively show the abnormal performance of the traffic,reflecting the advantages of visibility.The design based on transaction characteristics and transaction graph information shows good anomaly detection capabilities,provides a more comprehensive anomaly detection solution,is conducive to early perception of risks,responds,and effectively combats illegal and criminal activities using Bitcoin.