Tracing Information Flows Between Ad Exchanges Using Retargeted Ads
summary
Methodology builds controlled personas with an instrumented Chromium to visit 100 product pages per persona, collects 35,448 ad impressions (5,100 unique retarget ads) and uses Mechanical Turk labeling plus regex-based chain classifiers to identify four concrete sharing patterns: direct, cookie match, indirect and latent. Results detect both client- and server-side flows, show that 31% of cookie matches would be missed by prior heuristics, and reveal extensive server-side sharing inside Google after clustering domains.
AI-written summary · about
chapters (21)
- 0:00Intro
- 0:21Your Privacy Footprint
- 1:47Real Time Bidding
- 5:13Prior Work
- 5:28Challenge 1: Server Side Matching
- 6:12Challenge 2: Obfuscation
- 7:13Goal
- 7:36Key Insight: Use Retargeted Ads
- 8:17Contributions
- 9:08Using Retargets as an Experimental Tool
- 9:57Data Collection Overview
- 11:49Crowd Sourcing
- 12:57Final Dataset
- 13:39A look at Publisher Chains
- 14:38What is a Chain?
- 15:45Four Classifications
- 16:09Direct (Trivial) Matching
- 18:13Latent (Server-side) Matching
- 19:33Categorizing Chains
- 21:04Who is Cookie Matching?
- 22:59Summary
related talks
08mWhat Makes Us Human? A Tale of Advertising Fraud
01h 07mThe great data race ...
06mWho Am I? A Design Probe Exploring Real-Time Transparency About Online and Offline User ...
49mWhat Lies Beneath: What are they really tracking and how?
12mWhy am I seeing this: Democratizing End User Auditing for Online Content Recommendations
30mA Privacy Analysis of Cross-device Tracking
30mMeasurement in a Privacy First World
49mWeb Side Story: A Privacy Odyssey through Tracking and Fingerprinting