ADFRAUD : CLICK FRAUD DETECTION FOR MOBILE APPLICATION
1PG-Scholar, Department of CSE (Artificial Intelligence), JNTUA College of Engineering (Autonomous) Ananthapuramu, India.
2Associate Professor, Department of CSE, JNTUA College of Engineering (Autonomous) Ananthapuramu, India.
Mobile advertising is a crucial component in the mobile app ecosystem, with click fraud being a significant threat to its viability. This fraudulent activity, which includes ad clicks from malicious code or automated bots, undermines the ecosystem’s sustainability. Most current click fraud detection methods concentrate on examining ad requests from the server’s perspective. However, these methods can be easily bypassed, leading to a high rate of false negatives. Existing client-side (within the app) fraud detection divides tasks into two procedures: offline click request identification and an online process. In the offline stage, exact and probabilistic patterns are derived from URL tokenization, which then aid online click request identification and subsequent click fraud detection. This online detector is integrated into the app’s binary archive using binary instrumentation. A notable shortcoming of this method is its inefficiency in detecting click fraud and the latency resulting from the dual offline and online modes, negatively impacting user experience. Our proposed system offers an improved solution by introducing an efficient click fraud detection method on the server side. This new approach includes a pattern generation technique that accurately discerns between genuine and fraudulent ad requests. Implementing this server-side approach allows real-time fraud detection, covering aspects like fraud reviews, app downloads, and user comments. Consequently, latency issues are significantly reduced.
Keywords: Mobile advertising, Click fraud, Mobile app ecosystem, Malicious code, Automated bots, Server-side detection, False negatives, Client-side detection, Offline procedure, Online procedure, URL tokenization, Exact patterns, Probabilistic patterns, Binary instrumentation, Latency, User experience, Pattern generation, Real-time fraud detection.