CFDMA: A Novel Click Fraud Detection Method in Mobile Advertising

While the Cost-Per-click mechanism ensures the prosperity of the mobile advertising ecosystem, click fraud becomes a significant threat that results in huge losses for advertisers. On the one hand, some existing studies identify click fraud in a predefined test environment on the client side. However, these works neglect the influence of user interaction on fraudulent behaviours. On the other hand, some studies employ binary classifiers to detect click fraud on the server side without considering the particularity of the click fraud scenario. This paper presents the design and implementation of CFDMA, a click fraud detection system that conducts both server side and client side detection. CFDMA makes connections between fraudulent behaviours and explicit user inputs through observing sensitive system classes on the client side. CFDMA constructs the publisher graph on the server side and uses its information to detect invalid traffic. We evaluate the performance of CFDMA on real click datasets. Our evaluation demonstrates that CFDMA is capable of identifying invalid click traffic and fraudulent behaviours.