论文标题
发现指纹网络。使用基于行为的方法对浏览器跟踪的分析
Uncovering Fingerprinting Networks. An Analysis of In-Browser Tracking using a Behavior-based Approach
论文作者
论文摘要
近年来，互联网私人关系的重要性一直在不断上升。 [...]浏览器指纹是一种不需要cookie或持续标识符的技术。它从各种浏览器或设备属性中得出了足够唯一的标识符。学术工作已经涵盖了近十年的进攻和防御性指纹方法，观察到受欢迎程度的上升。本文探讨了互联网上浏览器指纹的当前状态。为此，我们实现FPNET - 一种基于FPMON的可扩展和可靠工具，通过观察其行为来识别大量网站上的指纹脚本。通过扫描Alexa前10,000个网站，我们发现了数百个同样行为脚本的网络。对于每个网络，我们确定其背后的演员。我们跟踪Google，Yandex，MaxMind，Sift或FingerPrintjs等公司，仅举几例。在三项互补研究中，我们进一步研究了有关i）随机化文件名或域的发现网络，ii）行为改变，iii）安全性。连续两次扫描表明，只有少于12.5％的页面不会更改脚本文件。通过我们基于行为的方法，我们成功地重新识别了近9,000个文件名或域更改的脚本，并且超过86％的脚本没有URL更改。安全分析表明，将TLS/SSL的采用至98％以上，特定的Web安全标头设置为30％以上的脚本。最后，我们对现代指纹识别的不可避免性及其对互联网用户隐私的影响表示担忧，因为我们认为许多用户不知道被指纹或没有足够的可能性来防止它。
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.
