AI Case Study

Eyeo leverages machine learning and object detection to develop ad blocking technology

The makers of Adblock Plus, Eyeo, are developing an ad detector based on machine learning and object detection. The technology, called Sentinel, aims at combating webpages that disguise ads from ads blockers and force ads on consumers. By using screenshots of Facebook feeds submitted by consumers, Eyeo is training the software to identify natively-served ads that would avoid normal ad blocking software.

Industry

Technology

Internet Services Consumer

Project Overview

"To optimize the training of the AI system, eyeo is inviting consumers worldwide to visit https://adblock.ai and submit screenshots of their ad-cluttered Facebook pages. Sentinel will learn to block Facebook ads first, and will then learn to recognize other ad types.

eyeo’s engineers developed this technology in reaction to aggressive tactics that attempt to re-insert ads onto pages for consumers who have ad blockers installed. Sentinel will look at the way an actual website is displayed on the screen to try and detect advertisements, instead of looking at the underlying code. This approach uses what is known as “object detection” in the machine learning world. The idea is that with merely a screenshot of a website, the AI will learn to identify where the ads on that screenshot are. Once that is achieved, likely through months of training, it can be deployed to automatically crawl the entire web, detecting and/or blocking ads." (businesswire)

Reported Results

Pilot; results not yet available.

Technology

"The neural network technology is currently being trained to detect ads and thwart circumvention technologies that try to evade consumers who use ad blockers." (businesswire)

Function

R And D

Product Development

Background

"Ad blockers of today rely heavily on a community of filter authors, who are continuously adding, changing and removing the filters which define what is blocked or what is considered an ad. It is quite a laborious task, often requiring constantly visiting the same websites again and again, just to make sure the stuff that is supposed to be blocked gets blocked, and stuff that should not — doesn’t." (Medium)

Benefits

Data

screenshots of Facebook feed