Source-linked AI summary
A First Look at Ad-block Detection: A New Arms Race on the Web
Muhammad Haris Mughees, Zhiyun Qian, Zubair Shafiq, Karishma Dash, Pan Hui
TL;DR
Publishers are increasingly countering ad-blockers, but the scale and technical approaches to ad-block detection were largely unknown. The paper measures Alexa top-100K sites using machine learning and characterizes detector strategies. It reports about 1,100 sites performing detection, with about 300 responding to blockers.
Problem
The scale and technical approaches of ad-block detection were largely unknown, despite publishers increasingly using detection to respond to ad-blockers.
Source: §1 · p. 1Method
The authors apply a machine-learning approach to Alexa top-100K sites and cluster detector strategies based on inserted JavaScript.
Source: §1 · p. 2Results
About 1,100 Alexa top-100K websites performed ad-block detection, and about 300 responded by asking users to disable blockers or blocking content.
Source: §6 · p. 11Takeaways & Limitations
The authors expect the arms race to continue as both ad-blockers and ad-block detectors adapt.
Source: §6 · p. 11The measurement is limited to sites that respond on the homepage without user input; sites requiring later navigation or content requests, and nonresponding detectors, were excluded.
Source: §3.3 · p. 5
Abstract
from arXiv · showhide
The rise of ad-blockers is viewed as an economic threat by online publishers, especially those who primarily rely on ad- vertising to support their services. To address this threat, publishers have started retaliating by employing ad-block detectors, which scout for ad-blocker users and react to them by restricting their content access and pushing them to whitelist the website or disabling ad-blockers altogether. The clash between ad-blockers and ad-block detectors has resulted in a new arms race on the web. In this paper, we present the first systematic measurement and analysis of ad-block detection on the web. We have designed and implemented a machine learning based tech- nique to automatically detect ad-block detection, and use it to study the deployment of ad-block detectors on Alexa top- 100K websites. The approach is promising with precision of 94.8% and recall of 93.1%. We characterize the spectrum of different strategies used by websites for ad-block detection. We find that most of publishers use fairly simple passive ap- proaches for ad-block detection. However, we also note that a few websites use third-party services, e.g. PageFair, for ad-block detection and response. The third-party services use active deception and other sophisticated tactics to de- tect ad-blockers. We also find that the third-party services can successfully circumvent ad-blockers and display ads on publisher websites.
1. INTRODUCTION
Online advertising supports free web content but also brings disruptive ads, tracking, and malvertising; publishers face growing ad-blocker use and have begun detecting it.
1. INTRODUCTION
- Online ads can disrupt content, track users across sites, and spread malware.Examples include autoplay videos, pop-ups, homepage takeovers, and aggressive profiling by ad exchanges.Source: §1 · p. 1
- Ad-blockers are a growing business threat to publishers, prompting some sites to detect blockers and refuse service while others pay for whitelisting.Users also adopt blockers for cleaner browsing; one cited study estimated 22% usage among the most active residential broadband users of a major European ISP.Source: §1 · p. 1
- The study addresses the unknown scale of ad-block detection and asks how many sites use it, which techniques they employ, and how blockers can counter it.The authors describe this as the first systematic characterization of the phenomenon.Source: §1 · p. 1
- The Alexa top-100K measurement identified about 300–1,100 detectors with a method reaching 94.8% precision and 93.1% recall.The study also found strategies ranging from passive detection to active deception.Source: §1 · p. 2
2. BACKGROUND
Ad-blockers remove ads or block web requests using filter rules, while publishers use page scripts to detect blockers and choose how to respond.
2. BACKGROUND
- Ad-blockers can remove page elements or block advertiser requests, using filter lists that specify domains and element selectors.The tools also filter network requests associated with profiling browsing behavior.Source: §2 · p. 2
- Publisher scripts monitor whether ads are visible and can respond with a mild notice or by blocking website content or functionality.Some responses ask users to disable blockers; aggressive responses can prevent access to any content.Source: §2 · pp. 2–3
- At the time of writing, the illustrated script could detect users with Adblock Plus.Source: §2 · p. 2
- In the VIPLeague example, the page loads, the blocker removes ads, and the detector then checks for them before showing a pop-up.The script waits 2,000 milliseconds before checking ad visibility and related element dimensions.Source: §2 · pp. 2–3, Figure 2 · p. 3
3. MEASURING AD-BLOCK DETECTION
The section builds a machine-learning approach that identifies ad-block detection from changes to page content with and without an ad-blocker. Random forests achieve the strongest reported classifier performance, while text and other feature types contribute complementary signals.
3.1 Overview
- Ad-block detection can add DOM nodes, alter existing-node styles, change text, replace page content, or redirect users to warning pages.The study assumes Adblock Plus's default configuration, which allows acceptable ads and may suppress some detections.Source: §3.1 · pp. 3–4
3.2 Methodology
- The method compares page loads with and without Adblock Plus, extracts content-change features, and trains a supervised classifier on labeled websites.Repeated unblocked loads help distinguish detector-related changes from dynamic page noise; the training set includes 200 positive samples and negatives from Alexa top 1,000 sites.Source: §3.2 · pp. 4–5
3.3 Feature Analysis
- The measurement excludes sites that respond only after specific content requests or navigation, and sites that detect ad-blockers without responding.Because the default Adblock Plus configuration permits acceptable ads, the authors also note that prevalence may be underestimated.Source: §3.1 · p. 3, §3.3 · p. 5
- Detection sites tend to change more text lines and add more div elements; text-based features have the highest information gain, exceeding 25%.Text features are followed by node- and style-based features in the information-gain ranking.Source: §3.3 · pp. 5–6, Table 2 · p. 5, Figure 5 · p. 6
3.4 Classifier Evaluation
- Random forests outperform C4.5 decision trees and naive Bayes, achieving 93.1% recall, 94.8% precision, and 96.0% AUC.Text features perform best alone, combinations improve classification, and all feature sets together achieve the best performance.Source: Table 3 · p. 6, §3.4 · p. 6
- The pruned decision tree uses word difference at its root, and its top three features come from different feature categories.The authors interpret this pattern as indicating that feature sets complement rather than duplicate one another.Source: Figure 6 · p. 6, §3.4 · pp. 6–7
4. AD-BLOCK DETECTION IN THE WILD
The study identifies ad-block detection across Alexa top-100K sites and examines how publishers implement it, from simple common-family scripts to more sophisticated third-party services. The analyzed strategies vary in how they detect ad-blockers and respond to them.
4. AD-BLOCK DETECTION IN THE WILD
- 1,089 Alexa top-100K websites were identified as detecting ad-blockers: 292 responded visibly, while 797 had detection scripts without visible behavior.The latter may reflect default-on acceptable ads in the Adblock Plus extension.Source: §4 · p. 7
4.1 JavaScript Collection
- A Chrome debugger collected parsed JavaScript snippets, including code unpacked through eval or added through iframe and script tags.The researchers used the parsed scripts to identify those responsible for ad-block detection.Source: §4.1 · p. 7
4.2 Clustering
- 292 detection snippets were represented by frequencies of 88 AST node types; PCA visualization showed a dense cluster and more distant outliers.The dense cluster was interpreted as similar approaches, while outliers represented customized and potentially more sophisticated approaches.Source: §4.2 · p. 7
4.3 Case Studies
- The dense cluster forms the “common family,” while manual inspection of outliers found third-party scripts including PageFair, FuckAdBlock, and Sourcepoint.Clustering was based on JavaScript-code similarity.Source: §4.3 · p. 7, §4.2 · p. 7
4.3.1 The Common Family
- Common-family detectors are usually 5–10 lines long and exploit ad-blockers’ obvious removal of ad frames.Some delay a one-time check at page load to give blockers time to act; checks commonly inspect ad-related DOM elements and their properties.Source: §4.3.1 · pp. 8–9
- Common-family responses range from requests to disable ad-blockers to removing content or redirecting users to warning pages.For example, one site preserves its content while another removes a content div after detection.Source: §4.3.1 · p. 9
4.3.2 PageFair
- PageFair appeared on 12 of the 292 detection websites and uses periodic checks, injected baits, and accessible browser-extension resources to detect ad-blockers.Its baits include hidden div, image, and script elements with ad-related names or URLs; the extension-resource checks cover eight popular ad-blockers.Source: §4.3.2 · p. 9
- PageFair replaces blocked Google ads with ads hosted on its EasyList-whitelisted domain; website behavior also changed when rechecked weeks later.The whitelist rule permits PageFair acceptable ads on partner websites.Source: §4.3.2 · p. 10
5. RELATED WORK
Prior research examined online advertising and ad-blockers, but the mechanisms of ad-block detection remained unstudied; this paper addresses that gap with a large-scale measurement.
Online Advertising
- Studies of online advertising examined malvertising, including its prevalence and evasion techniques across websites.Li et al. studied malvertising across 90,000 websites; Zarras et al. measured more than 60,000 ads and found around 1% exhibited malicious behavior.Source: §5 · p. 10
- Prior advertising research also measured web tracking and mobile ad targeting, including the prevalence of tracking services and users being tracked.Roesner et al., Metwalley et al., and Nath studied tracking or ad targeting in web and mobile settings.Source: §5 · pp. 10–11
Ad-blockers
- Ad-blocker research measured adoption and acceptable-ad whitelists, and proposed automatically blacklisting intrusive ads.The cited work also studied whitelist contents and developed a classifier to identify advertising and tracking services.Source: §5 · p. 11
5. RELATED WORK
- No prior work had studied ad-block detector mechanisms, motivating this paper’s large-scale measurement of ad-block detection.The authors describe this as the first large-scale measurement study of ad-block detection.Source: §5 · pp. 10–11
6. CONCLUSION
The study estimates ad-block detection across Alexa’s top 100K websites and describes how sites and third-party detectors respond. It reports that detection and detector strategies are already present, and anticipates continued adaptation in the arms race.
6. CONCLUSION
- About 1,100 Alexa top-100K websites perform ad-block detection, and about 300 respond by asking users to disable blockers or restricting content.The study also reports increasing use of third-party ad-block detectors such as PageFair.Source: §6 · p. 11