Detection of Fraud in Mobile Advertising using Machine Learning

Abstract

With ongoing advancements in the field of technology, mobile advertising has emerged as a platform for publishers to earn profit from their free applications. An online attack commonly known as click fraud or ad fraud has added up to the issue of concerns surfacing mobile advertising. Click fraud is the act of generating illegitimate clicks or data events in order to earn illegal income. Generally, click frauds are generated by infusing the genuine code with some illegitimate bot, which clicks on the ad acting as a potential customer. These click frauds are usually planted by the advertisers or the advertising company so that the number of clicks on the ad increases which will give them the ability to charge the publishers with a hefty sum per number of clicks. A number of studies have determined the risks that click fraud poses to mobile advertising and a few solutions have been proposed to detect click frauds. The solution proposed in this paper comprises of a social network analysis model – to detect and categorize fraudulent clicks and then test sample datasets. This social network analysis model takes into consideration a wide range of parameters from a large group of users. A detailed study is conducted for analyzing these parameters in order to separate the parameters, which affect the click fraud generation process largely. These parameters are then tested and categorized into sample datasets. The mobile advertising industry forms a large part of the revenue generated by the advertising industry. Hence, detection of click fraud in mobile advertising is important to ensure that no illegitimate sources are used to generate this revenue. To be precise, the proposed method touches an accuracy of about 92%.