Adaptive detection for dynamically changing ad-fraud triggers

Investigate whether advanced ad fraud requires adaptive detection methods and, if so, determine which types of adaptive methods are needed to detect ad networks that dynamically adjust their triggering conditions.

Background

The paper detects shake-to-trigger semi-drive-by splash ads using threshold values from the TC260 advertising standard. The authors note that lower thresholds can increase false positives, whereas higher thresholds can increase false negatives, and that ad networks may adaptively change their triggering conditions to evade static detection rules.

The unresolved problem is whether adaptive ad-fraud behavior necessitates adaptive detection and, contingent on that determination, what detection techniques can remain effective against dynamically changing trigger conditions. The question is left for future research rather than resolved by the AdHive evaluation.