Placement Fraud Detection On Smart Phones:A Joint Crowdsourcing And Data Analyzing Based Approach

With the widespread use of mobile devices,mobile online advertising is taking more and more market share.Cost per click and cost per view are the most popular pricing modes in mobile internet advertising,which take effective clicks or displaying duration as the charging basis.However,at the same time,click fraud,which uses illegal and invalid clicks to fraud advertisers in order to obtain unreasonable income,become a serious problem.Most of the previous studies on click fraud in websites focused on network traffic data analysis.These studies cannot solve the placement fraud problem,which use invalid placement to mislead user to click on it in mobile apps.In this paper,we propose a joint crowdsourcing and data analyzing based placement click fraud detection system.For the characteristic of placement fraud in mobile apps,automatic processing cannot cover every possible fraud.To overcome this,our report system provides a platform to find all possible placement fraud through crowdsourcing.Report system has three main services: a monitor service for monitoring user's call;a layout service for recording the screen;a data service for recording the backend data.Because the placement fraud only appears when users use the apps,the report system based on crowdsourcing can cover every possible placement fraud.We implement our system in 10 tablets with500 apps to evaluate its effectiveness.Experiment result shows that our approach can record enough data to analysis which app has placement fraud.What's more,our system can figure out some special placement fraud which pop ads when user is using other apps.This placement fraud cannot be solved through automatic method in previous studies.