DECAF: detecting and characterizing ad fraud in mobile apps

  • Authors:
  • Bin Liu;Suman Nath;Ramesh Govindan;Jie Liu

  • Affiliations:
  • University of Southern California;Microsoft Research;University of Southern California;Microsoft Research

  • Venue:
  • NSDI'14 Proceedings of the 11th USENIX Conference on Networked Systems Design and Implementation
  • Year:
  • 2014

Quantified Score

Hi-index 0.00

Visualization

Abstract

Ad networks for mobile apps require inspection of the visual layout of their ads to detect certain types of placement frauds. Doing this manually is error prone, and does not scale to the sizes of today's app stores. In this paper, we design a system called DECAF to automatically discover various placement frauds scalably and effectively. DECAF uses automated app navigation, together with optimizations to scan through a large number of visual elements within a limited time. It also includes a framework for efficiently detecting whether ads within an app violate an extensible set of rules that govern ad placement and display. We have implemented DECAF for Windows-based mobile platforms, and applied it to 1,150 tablet apps and 50,000 phone apps in order to characterize the prevalence of ad frauds. DECAF has been used by the ad fraud team in Microsoft and has helped find many instances of ad frauds.