Section: New Results

Large-scale measurement of invisible images for Web tracking

In joint work with Arnaud Legout (DIANA team, Inria Sophia Antipolis), we perform large scale Web measurements to evaluate Web tracking and privacy leaks in every-day Web browsing. Unlike the related work, our study focuses on the third-party HTTP requests for invisible images. We have identified two types of images that are invisible to the end user and most likely used for Web tracking: one-pixel images and empty images.

We have visited 4,351,318 pages from 38,000 web sites and identified that almost half of the third-party images are invisible to the end user. This finding raises a lot of concerns regarding Web tracking and user privacy on the Web. We made the first evaluations on how much of this invisible tracking is prevented by the popular browser extensions used for the privacy protection. We also find the top invisible trackers and the invisible trackers not blocked by the browser extensions.

We continue this work by analysing all the HTTP requests and responses that lead to invisible images in order to (1) provide a fine-grained classification of third-party cookie tracking; (2) analyse new techniques of cookie-synching used by the companies; (3) evaluate redirection chains that lead to user's information exchange between various companies; (4) identify companies that use invisible images for none of the known tracking techniques, and analyse such requests and responses further to reveal new Web tracking technologies.