Flow Extraction and CC detection in HPC

Flow-based detection of network covert channels typically requires the extraction of flows from traffic recordings. In real, traffic-heavy networks, these recordings quickly result in large files. Processing such large files presents new challenges that can be dealt with by applying high-performance computing (HPC) programming models and resources, a use case that has been mostly overlooked so far. This thesis addresses this gap by presenting a proof-of-concept implementation of flow extraction in HPC environments, including a benchmark to evaluate performance and efficiency and identify bottlenecks. Insights from the development and benchmarking are presented to contribute to future work on the subject.

A usability and user study for internet censorship

As internet censorship can be a threat to society, it is important to raise public awareness of this issue. Compared to the technical aspects of internet censorship, the educational aspect is relatively under-researched. To fill this gap, an exhibit was done to demonstrate the effects of internet censorship in the world. The aim of this thesis is to conduct a usability study of this exhibit, as well as a user study of the general population, in order to optimize the effectiveness of censorship education.

Perception of Censorship in Large Language Models

Large language models do not always answer in full. Sometimes they refuse openly, but they can also leave information out without any notice, so the shortened answer looks complete. In my bachelor thesis, I am designing a user study on whether people notice such omissions, whether they suspect censorship or simply missing knowledge behind them and how this affects their trust in the model.