Modern fraud trends in RTB
summary
Company processes ~6 billion daily requests and finds mobile apps now near 50 percent of traffic so app fraud surges. Detection uses a three stage waterfall filtering: simple rule blocks (catching about 8 percent of web fraud), blacklists (2–3 percent) and ML (2–3 percent). For apps effective heuristics include a sleep-time metric that detects background ad pumping, entropy scoring to find one‑user‑per‑app patterns, geolocation flapping and non‑store provenance checks. Simple rules already remove roughly 70 percent of obvious fraud with low false positives; Charlie Proxy used for request inspection.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:57Mobile traffic and fraud growth
- 2:59Fraud classification and types
- 8:12Web traffic fraud prevention
- 10:23App fraud specifics
- 13:27Review analysis and motivated traffic
- 21:09App fraud detection rules
- 26:09Geolocation and device identification
- 29:45Summary and development plans
- 30:50Q&A
related talks
36mFraud in mobile applications: how to define and detect
19mWhich mobile fraud solution is right for you?
37mFraud prevention in the age of online privacy
37mPreventing Fraud @ Grab
16mProtecting your application from bot attacks (PRT019)
40mReversing Fraudulent Apps
30mFair game: the ethics of telco fraud
35mBreaking fraud & bot detection solutions