AdGraph: A Graph-Based Approach to Ad and Tracker Blocking

Explore a graph-based machine learning approach for detecting and blocking online advertising and tracking resources in this IEEE conference talk. Learn about AdGraph, a novel method that builds a graph representation of HTML structure, network requests, and JavaScript behavior to train a classifier for identifying ads and trackers. Discover how AdGraph's multi-faceted approach makes it less susceptible to evasion techniques compared to existing solutions. Examine the evaluation results on Alexa top-10K websites, showcasing AdGraph's high accuracy in replicating human-generated filter lists and its ability to identify mistakes in these lists. Gain insights into AdGraph's implementation as a Chromium modification, its performance impact on page loading, and its effectiveness compared to stock Chromium and AdBlock Plus. Understand the potential of AdGraph as an accurate and performant solution for online ad and tracker blocking.

Intro
Online Advertising
What are Ads and Trackers?
Filter List Based Blocking
Machine Learning Based Blocking
Outline
Cross-layer Context
JavaScript Attribution
Chromium Instrumentation
Features Extraction
Evaluation: Accuracy
Evaluation: Performance
Key Takeaways
Questions?