Behavioural Feature Analysis for Anomalous Click Detection in Mobile Advertising Environments: Toward In-App Browser-Specific Detection

Abstract

Mobile in-app browsers (IABs), implemented through WebView components, have become a dominant channel for rendering advertising content within mobile applications. This environment presents distinctive challenges for ad fraud detection due to its constrained JavaScript execution context, opaque rendering pipeline, and limited access to conventional browser-level signals. This paper presents a behavioral feature analysis targeting anomalous click detection in mobile advertising environments, with particular attention to in-app browser (IAB) deployment contexts. A multi-dimensional feature taxonomy is proposed, organized across four analytical dimensions: temporal patterns, gestural characteristics, device fingerprinting, and network-level attributes. Using three publicly available datasets—TalkingData AdTracking (approximately 200 million click records), FDMA 2012 BuzzCity, and the labeled real-time bidding (RTB) logs reported in recent literature—this study evaluates the discriminative capacity of each feature dimension for characterizing click behavior anomalies indicative of fraudulent activity. Because the publicly available datasets used in this study originate from general mobile advertising platforms rather than IAB-specific instrumentation, the results are interpreted as evidence for mobile ad fraud detection that is directionally—but not conclusively—applicable to IAB environments. The analysis indicates that temporal aggregation features and device consistency metrics provide the strongest detection signals across the evaluated mobile advertising datasets. A discussion on the trade-off between detection granularity and user privacy constraints under data minimization principles is also provided. The findings contribute empirical guidance for practitioners developing fraud mitigation strategies within mobile advertising environments, including IAB contexts.