On the Impact of Internal Webpage Selection when Evaluating Ad Blocker Performance

Philip Gunnarsson, Adam Jakobsson and Niklas Carlsson

Paper: Philip Gunnarsson, Adam Jakobsson and Niklas Carlsson, "On the Impact of Internal Webpage Selection when Evaluating Ad Blocker Performance", Proc. IEEE International Symposium on the Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS), Nice, France, 2022. (pdf)

Abstract: Not all ad blockers achieve the same blocking success and, depending on their implementation, they can either improve or hurt the web performance experienced by users. Borgolte and Feamster (2020) recently provided the first extensive evaluation of how privacy-focused browser extensions affect a user's web performance. However, while their work provides a nice comparison of the performance impact that different extensions may have, their evaluation only considered landing pages of a single set of websites. In this paper, we focus specifically on performance comparisons when considering different sets of webpages. For example, we study the impact of whether a page is a landing page or an internal page, whether a page is popular or less popular, as well as the impact of in which country/region the company registering the website is operating (used as a proxy for the primary target market). For our evaluations, we use pairs of webpages carefully selected from the recently proposed Hispar list (Aqeel et al. 2020) and compare the performance of the most popular blocking extensions (Adblock Plus, uBlock Origin, Ghostery, and Private Badger) considered by Borgolte and Feamster with a baseline case in which we do not use any extension. While we observe clear differences in the distribution statistics of the metrics considered, several observations were consistent across all dimensions, including whether we consider landing pages or internal pages. The paper highlights some of these invariants and discusses their implications. In addition, our measurements (and the differences observed by different ad-blockers) also reveal new insights into how internal vs. landing pages of different webpage categories (e.g., based on popularity or region) differ in their composition and resource usage.

Data collection tool and datasets

To help build upon our work, below, we make available the dataset.

The dataset can be downloaded here.

Citation format

When using the dataset, please cite the conference version of the paper:

Philip Gunnarsson, Adam Jakobsson and Niklas Carlsson, "On the Impact of Internal Webpage Selection when Evaluating Ad Blocker Performance", Proc. IEEE International Symposium on the Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS), Nice, France, 2022. (pdf)