Fraud in mobile applications: how to define and detect
summary
Two practical detectors for mobile-app RTB fraud: a background ad detection metric that measures each app–device pair’s maximum silent interval (t_max) and flags continuous request streams with implausibly small gaps (example cutoff ~6 hours), and an entropy per app score using Shannon entropy across users to spot apps whose traffic concentrates on few devices (rewarded/stimulated apps). Offline batch jobs compute signals, a cached model enforces real‑time blocking; deployed system drops ~15% of bad traffic with the background rule and adds ~1.5–2% via entropy, overall ~15–20% filtered.
AI-written summary · about
chapters (9)
- 0:00Fraud introduction
- 1:35Mobile apps as new segment
- 3:00Classic fraud types
- 5:32Fraud in mobile apps
- 7:39Method 1: Background traffic
- 15:06Method 2: Low traffic entropy
- 21:52Comparing methods
- 24:43Conclusion and outlook
- 25:46Q&A
related talks
45mModern fraud trends in RTB
19mWhich mobile fraud solution is right for you?
37mFraud prevention in the age of online privacy
13mBehavioral Analytics for Real-Time Fraud Prevention
35mBreaking fraud & bot detection solutions
42mOutlier detection methods for detecting cheaters in mobile gaming
37mPreventing Fraud @ Grab
30mFair game: the ethics of telco fraud