Title:

Statistical Analysis and Classification of Real-Time Network Communication to Detect Web-Trackers

Description:

Third-party trackers are in a continuous rise to track user activities because data is the fuel for analytics or Ad tracking. Privacy violation by trackers has become a serious problem and the most effective ways to block tracker are based on using blacklists. These blacklists are effective but comes with limitations. New techniques need to be explored to be able to identify if a request is made to a tracker or a non-tracker. In this master thesis, a different approach to classify if a request is made to a tracker or a non-tracker based on the features of an HTTP request. On exploration of hundereds of HTTP requests we found that characteristics of an HTTP request varies from a tracker to a non-tracker. These distinct characteristics have been used with the help of statistical analysis and machine learning to determine the difference between trackers and non-tracker requests. Both supervised and unsupervised learning was used to determine if the requests were made to tracker or non-tracker. The accuracy of supervised learning was far better than unsupervised learning. Although the number of false positives of the results is high, it discards the usage of block lists.

Supervisor: Philip Raschke [1]

Type:  Master Thesis

Duration: 6 months