Abnormal Node Classification and Security Detection for Cross-border SME E-commerce Using Blockchain Network Topology Algorithms

Abstract

In light of the pressing concerns regarding the inadequacy of transaction security, efficiency, and transparency within the financial system, this study endeavors to enhance the security of digital economy transactions for small and medium-sized enterprises engaged in cross-border e-commerce through the application of blockchain network topology algorithms. Specifically, the research introduces an innovative approach to classifying abnormal nodes, leveraging a dynamic update algorithm rooted in blockchain network topology. Additionally, it proposes a method for detecting security in digital economy transactions, also grounded in blockchain network topology algorithms. Under the conditions of a total of 60,000 records of real transactions in Bitcoin and Ethereum and a node scale of 100 to 1,000, the experiment uses a combination of cosine and Euclidean distance to calculate the transaction frequency, amount and time series characteristics of nodes and complete clustering. Subsequently, a sliding time window is used to dynamically update the node similarity threshold to identify anomalies. Compared with the three benchmark methods of density clustering, graph convolutional network and autoencoder, the proposed blockchain network topology algorithm has a root mean square error of 0.09, a mean absolute error of 0.09, an anomaly detection accuracy of 8.6%, and a transaction success rate of 1.1%, which is jointly determined by a 1.8-millisecond delay and a throughput of 13.2 transactions per second. All indicators are superior to the benchmark methods. The blockchain network topology algorithm can significantly improve transaction security and system stability, which is of great significance for promoting sustainable economic growth and social stability.