Fraudulent traffic detection in online advertising with bipartite graph propagation algorithm

作者：

Highlights：

• We propose novel graph models to detect fraudulent IPs in online ad impression.

• We consider behavioral patterns of IP fraud: power-law distribution and similarity.

• We integrate the novel graph models with the random forest algorithm.

• Our model outperforms the basic bipartite LPA graph model in precision by 17.9%

摘要

•We propose novel graph models to detect fraudulent IPs in online ad impression.•We consider behavioral patterns of IP fraud: power-law distribution and similarity.•We integrate the novel graph models with the random forest algorithm.•Our model outperforms the basic bipartite LPA graph model in precision by 17.9%

论文关键词：Online advertising,Fraudulent ad traffic,Fraud detection,Bipartite graph,Label propagation

论文评审过程：Received 18 April 2020, Revised 26 March 2021, Accepted 6 July 2021, Available online 12 July 2021, Version of Record 23 July 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115573