Graph-based Malicious Account Groups Detection in Electronic Trading of Bulk Commodities

Xiang Li,Yihao Li,Ping Yi

DOI: https://doi.org/10.1109/cscloud-edgecom54986.2022.00020

2022-01-01

Abstract:Bulk commodities, such as food and minerals, play a vital role in social life, but there are many risks in electronic trading of bulk commodities, especially the malicious account groups composed of a large number of malicious accounts which try to manipulate market prices through continuous trading. The key to preventing this risk lies in the timely and accurate detection of malicious account groups. However, the detection of malicious account groups is a difficult task in electronic trading of bulk commodities due to the flexibility of electronic transactions. Traditional methods rely on human review. But it has low efficiency and low accuracy. This article propose a new approach which utilizes the graph theory and graph embedding to detect the malicious account groups hidden in the market. Graph theory is used to build a comprehensive trading network with the help of transaction records. Graph embedding theory is applied to extract latent features of the accounts so as to pick out the malicious accounts and find the groups. This graph-based methodology makes it possible to find the malicious account groups lurking in the market automatically, as well as improve the accuracy and efficiency remarkably.