基于链接预测匹配的动态社交机器人检测方法OA

The malicious social bots in online social networks threaten the information security and af-fect the credit system of social networks,and it is imperative to effectively detect bot accounts.To ad-dress the problem that existing bot detection methods are difficult to portray the dynamic link behavior in social graphs,a dynamic bot detection method based on link prediction is proposed in this paper.We build node link prediction models based on the known links of bots and human users respectively,to learn the difference between the link behavior of both types of nodes.Finally,the classification is achieved by examining the agreement of unknown node link behavior with respect to two types of link prediction models.On the Twitter dataset,the detection accuracy is significantly improved compared with the baseline method without using temporal data,which verifies the effectiveness of this method in using temporal data.