Alicja Martinek’s Publication Recognized by ACM on the Kudos Platform

From November 10–14, 2025, in Seoul, the 34th edition of the ACM International Conference on Information and Knowledge Management (CIKM) took place. This prestigious event attracts researchers from around the world who focus on processing information and transforming it into knowledge.

As part of the conference, the PhD Symposium was held, giving doctoral students the opportunity to present their latest research and receive valuable feedback from experienced experts.

Kudos is an international platform that helps researchers promote their work and increase its visibility. It enables presenting publications in an accessible format, adding summaries and graphics, and tracking the reach and impact of scientific work.

Alicja Martinek, representing the Data Science Department, presented her paper titled:
“Real until proven fake – Source-Level Audio Deepfake Detection (with PIPNet)”.

This research focuses on classifying generators used for synthetic speech production. The novelty of this work lies in an image-based approach that detects artifacts left by such generators in audio files. The use of the PIPNet network makes it possible to identify “prototypes” that indicate membership in a specific generator class. Thanks to this, we can see what cannot be heard!

The results obtained will serve as a foundation for further research within projects conducted by the Audio-Visual and Biometric Analysis Department. The presentation at the conference was very well received, and the discussions initiated with mentors proved to be highly inspiring.