Brave Proposes a Machine Learning Approach for Ad Blocking

This week, Brave unveiled new research that is under submission to an upcoming conference regarding how to improve and automate ad blocking with AdGraph, a graph-based machine learning approach for detecting ads and trackers on a given web page.

AdGraph alleviates the need for manual filter list curation by using machine learning to automatically identify patterns in the page load process to block ads and trackers. AdGraph automatically and effectively blocks ads and trackers with 97.7% accuracy. AdGraph even has better recall than filter lists, as it blocks 16% more ads and trackers with 65% accuracy. The analysis also shows that AdGraph is fairly robust against adversarial obfuscation by publishers and advertisers that bypass filter lists.

Brave’s Chief Scientist, Dr. Ben Livshits, worked with Peter Snyder, a privacy researcher at Brave, and researchers from the University of Iowa (Umar Iqbal and Zubair Shafiq) and the University of California Riverside (Shitong Zhu and Zhiyun Qian) on this project. The full paper can be downloaded from here. The team is looking at deploying these techniques within Brave over time.

The team explained that filter lists are widely deployed by ad blockers to block ads in web browsers; however, these filter lists are manually curated based on informal crowdsourced feedback, which brings a number of maintenance challenges. AdGraph addresses these challenges with an approach that relies on information obtained from multiple layers of the web stack (HTML, HTTP, and JavaScript) to train a machine learning classifier to block both ads and trackers.

click image to enlarge

With these three ingredients brought together, the researchers showed that they can train supervised machine learning models to automatically block ads and trackers. The team also noted that more and more financially motivated publishers and advertisers are expected to employ adversarial obfuscation techniques to evade ad blockers.

Because crowdsourced filter lists used by state-of-the-art ad blockers can be easily evaded using simple obfuscation techniques, AdGraph’s resistance to those obfuscation attempts by publishers and advertisers represents an important technical advancement in the rapidly escalating ad blocking arms race.

Related articles

Ready for a better Internet?

Brave’s easy-to-use browser blocks ads by default, making the Web cleaner, faster, and safer for people all over the world.