Uncovering Fingerprinting Networks. An Analysis of In-Browser Tracking using a Behavior-based Approach

TL;DR

This paper presents FPNET, a behavior-based tool for identifying browser fingerprinting scripts on websites, revealing widespread fingerprinting networks and raising privacy concerns.

Contribution

The study introduces FPNET, a scalable, reliable method to detect fingerprinting scripts based on behavior, and provides a comprehensive analysis of fingerprinting networks on top websites.

Findings

Identified several hundred fingerprinting networks on top websites.

Successfully re-identified 86% of scripts despite URL changes.

Most fingerprinting scripts use TLS/SSL and security headers.

Abstract

Throughout recent years, the importance of internet-privacy has continuously risen. [...] Browser fingerprinting is a technique that does not require cookies or persistent identifiers. It derives a sufficiently unique identifier from the various browser or device properties. Academic work has covered offensive and defensive fingerprinting methods for almost a decade, observing a rise in popularity. This thesis explores the current state of browser fingerprinting on the internet. For that, we implement FPNET - a scalable & reliable tool based on FPMON, to identify fingerprinting scripts on large sets of websites by observing their behavior. By scanning the Alexa Top 10,000 websites, we spot several hundred networks of equally behaving scripts. For each network, we determine the actor behind it. We track down companies like Google, Yandex, Maxmind, Sift, or FingerprintJS, to name a few. In three complementary studies, we further investigate the uncovered networks with regards to I) randomization of filenames or domains, II) behavior changes, III) security. Two consecutive scans reveal that only less than 12.5% of the pages do not change script files. With our behavior-based approach, we successfully re-identify almost 9,000 scripts whose filename or domain changed, and over 86% of the scripts without URL changes. The security analysis shows an adoption of TLS/SSL to over 98% and specific web security headers set for over 30% of the scripts. Finally, we voice concerns about the unavoidability of modern fingerprinting and its implications for internet users' privacy since we believe that many users are unaware of being fingerprinted or have insufficient possibilities to protect against it.