Analysis of mouse-based identification on the web

OData support
Dr. Gulyás Gábor György
Department of Automation and Applied Informatics

In case of many web-based, online companies, it frequently happens that their business model is based on the practice of intensive and expansive data collection. Several methods have spread to realize this in practice, all of which keep track of how people use the web. Quite a few of these privacy-infringing activities could have been avoided due to their technology, just as the most widespread tracking method called tracking cookies can also be avoided by deleting them from time to time.

Since it has become easy to deal with cookie-based tracking, the purpose of this thesis is to examine technologies in which the defence of privacy can be realized with less ease. The topic is similar to that of my BSc thesis, entitled ’Analyzing mouse heatmaps of web users’ and the focus now is also related to the examination of identification based on mouse activities. The aim of this study is to explore this area as regards of privacy and to examine identification, which has been conducted by studying users’ mouse movements from different aspects based on different features.

Firstly, the literature is reviewed, where, having presented a few existing tracking methods, it is also examined to what extent characteristic, personalized attributes can be connected to users based on their mouse activities, that is whether it is possible to identify their behaviour in a biometric way. If unique, individual features can be found, by which users can be differentiated, which means they can be uniquely identified as well. These activities could also be used for profiling.

To perform this examination, data had to be collected. To realize it, I created a Firefox extension to record mouse movement data, which process had also been prepared and the users had been supplied with the necessary information concerning the collection. The framework in which the data were recorded had been written in JavaScript and PHP languages.

Finally, 23 users participated in the collection of data. The evaluation of the data was accomplished by Python programs automatically, with the help of functions formed according to different aspects, since the dataset consisted of 1.2 million records. The results of the examinations reinforce the original hypothesis, which claims that feature attributes characteristic of the individual can be extracted which help identification on the basis of the users’ mouse movements.


Please sign in to download the files of this thesis.