Cooperative malicious network behavior recognition algorithm in E-commerce

Abstract

Because of the virtuality of the e-commerce transaction, its transaction behavior faces the serious security threats. However, the existing recognition algorithms of malicious transaction behaviors only rely on traditional malicious features, resulting in low recognition efficiency. This paper proposes a novel technique for identifying collaborative malicious network behavior in e-commerce: the quantum behavior-based particle swarm optimization method with global search capability, which is combined with the rapidly converging k-means method to improve efficiency and effectiveness of the algorithm. The paper first illustrates that the current malicious network behavior is diversified and collaborative, it is necessary to identify malicious network behavior from a higher level-cooperation level, and then proposes original cooperative malicious behavior features of e-commerce. Second, the use of information transfer entropy (ITE) based on quantum potential energy is introduced to designate the node with the largest ITE in the cluster as the cluster center, avoiding randomly initializing cluster center nodes and calculating the distance repeatedly. Third, the existence of collaborative malicious network behavior and the feasibility of the algorithm are proved using the fixed point theorem in the probability metric space. Meanwhile, quantum behavior-based particle swarm optimization is used to extract cluster centers as a means of improving the spatial distribution and collaborative state of particle nodes, and enables the clustering algorithm to have global search ability. The simulation results demonstrate that the our schemes improve the efficiency and accuracy of identifying cooperative malicious network behavior in e-commerce recognition.