Sandra Bergmann

Sandra Bergmann, M. Eng.

Department of Computer Science
Chair of Computer Science 1 (IT Security Infrastructures)

Room 12.125
Martensstr. 3
91058 Erlangen

Phone number: +49 9131 85-67649
Fax number: +49 9131 85-69919
Email: sandra.daniela.bergmann@fau.de
Website: https://www.cs1.tf.fau.de/person/sandra-bergmann/

About me

Hi, I’m Sandra Bergmann, before I joined the FAU I received my M. Eng. degree from the Technische Hochschule Georg-Simon Ohm Nuermberg, Germany. Since December 2021 I am a PhD candidate in the research group Multimedia Security at the IT Security Infrastructures Lab. My research interests include AI compression, image forensics and deepfake detection.

Recent Projects

Below a short description of my current project. I recently focus on AI compression forensics, because of the new upcomping standard JPEG AI. To build new reliable image forensic tools for this new compression format it is important to forensic analyze AI compressed images to get a first insight what traces AI compression leaves in the images. Some details can be seen below:

Frequency-Domain Analysis of Traces for the Detection of AI-based Compression
We have done the first systematic analysis of AI-compressed images. AI-based compression leaves traces in the frequency domain, which can be used to detect AI compressed images. In our work we investigate these traces and have performed various experiments on the detectability, robustness and repeated compression of the artifacts in the frequency domain.

Impact of AI-based Codecs on Biometrics for Law Enforcement
We have investigated how AI compression impacts iris, fingerprint and soft-biometrics (fabrics and tattoo) images for different AI codecs. The influence of AI compression on biometric images depends on the special characterisitics of its data and on the loss function of an AI compressor.

Publications

2024

Bergmann, S., Moussa, D., Brand, F., Kaup, A., & Rieß, C. (2024). Forensic analysis of AI-compression traces in spatial and frequency domain. Pattern Recognition Letters.
URL: https://www.sciencedirect.com/science/article/pii/S0167865524000503
Moussa, D., Bergmann, S., & Rieß, C. (2024). Unmasking Neural Codecs: Forensic Identification of AI-compressed Speech. In ISCA (Eds.), Proc. Interspeech 2024 (pp. 2260-2264). Kos, GR.
URL: https://www.isca-archive.org/interspeech_2024/moussa24_interspeech.html#

2023

Bergmann, S., Moussa, D., Brand, F., Kaup, A., & Rieß, C. (2023). Frequency-Domain Analysis of Traces for the Detection of AI-based Compression. In IEEE (Eds.), 2023 International Workshop on Biometrics and Forensics (IWBF). Barcelona, ES.
URL: https://faui1-files.cs.fau.de/public/publications/mmsec/2023-Bergmann-IWBF.pdf