MÃ¼ller, T. (2015):
On detecting Web-Tracking
Web-Tracking is an incredible big business focusing on the collection of as many user data as possible and use it -- among other scenarios -- for hand-crafted advertising. Trackers are constantly seeking to improve their tracking mechanisms in order to be able to gather more user-based data. The most recent developments of this research are fingerprinting techniques: Browser- and Canvas-Fingerprinting. Fingerprinting techniques differ to the de facto standard for implementing Web-Tracking -- Cookies -- in several aspects. Firstly, it is very hard to notice that a web site is fingerprinting the user. Secondly, even if the user knows that he is being fingerprinted, it is very hard to effectively block it. Lastly, fingerprinting techniques enable the trackers to recognize users, which consequently leads to even more user-based data, because this enables them to track users among different web sites. Using the collected user data facilitates the trackers to create detailed user profiles, which clearly threatens the privacy of users. In order to increase privacy being endangered by Web-Tracking, the first step required is to be able to detect the use of fingerprinting techniques. This is the basic task this thesis sets out to solve. The detection part is realized using proven detection mechanisms from different fields like Data Mining or Knowledge Discovery in Databases. The results of the thesis show that Canvas-Fingerprinting is reliably detectable with each of those classifiers. The more general Browser-Fingerprinting is way harder to detect, but the best classifiers still managed to achieve a very high success rate. Consequently, it is safe to say that the developed system is capable of detecting the use of fingerprinting techniques on a web site and can therefore serve as a first step towards a system which can be used to increase privacy by mitigating fingerprinting techniques.