Blockchain Abnormal Transaction Detection Based on Deep PCA and Bayesian Optimization
Due to the complex trading scenarios and rich trading modes of blockchain,blockchain transactions are frequently threatened by illegal behaviors such as anonymous attacks,Ponzi schemes,phishing attacks,etc.These abnormal behaviors cause huge economic risks to the development of smart grid based on blockchain technology.Aiming at the problem of poor comprehensive performance of blockchain abnormal transaction detection,the characteristics of high data dimensions and imbalanced positive and negative samples of transaction data are analyzed.A blockchain abnormal transactions detection method based on deep principal component analysis(PCA)and Bayesian optimization is proposed.By designing the deep PCA model,linear and nonlinear dimensionality reduction of blockchain transaction data is achieved.The Bayesian optimization is employed to optimize the random forest hyper-parameters,and the optimized random forest classifier is utilized to effectively solve the problem of unbalanced positive and negative samples.And finally abnormal transactions are detected in the blockchain.The experiments based on Elliptic and power grid blockchain transaction datasets show that the proposed method improves the comprehensive performance of blockchain abnormal transaction detection.
国家科技期刊平台
NSTL
万方数据
