Bayesian Source Attribution of Synthetic Images

Vaidya AM, Bergmann S, Riess C (2026)

Publication Type: Conference contribution

Publication year: 2026

Publisher: Gesellschaft fur Informatik (GI)

Book Volume: P-379

Pages Range: 141-152

Conference Proceedings Title: Lecture Notes in Informatics (LNI), Proceedings - Series of the Gesellschaft fur Informatik (GI)

Event location: Hamburg DE

DOI: 10.18420/sicherheit2026_12

Abstract

The authenticity of digital images has become increasingly important, particularly for the question whether an image is photographic (“real”) or generated by an AI system. Also the provenance of an AI-generated image may be an important cue in a forensic analysis. The task of deepfake source attribution aims to distinguish K + 1 classes, namely whether an image is real or generated by one out of K image generators. This work proposes a method for source attribution with a Bayesian fusion of binary detectors with zero-shot priors. Priors from three recently proposed zero-shot statistics are compared, namely the LPIPS distance, latent cosine similarity, and coding-cost gaps. Our results show that the accuracy in a 8-class scenario increases about 8.1% compared to the baseline detector. Furthermore, the false negative rate for a binary detection task of generated images increases by 10.6%, which may aid applications like fraud detection. Code is available at https://github.com/vaidya-apoorva/bayesian-ai-image-attribution.

Authors with CRIS profile

Sandra Bergmann Lehrstuhl für Informatik 1 (IT-Sicherheitsinfrastrukturen) Christian Rieß Lehrstuhl für Informatik 1 (IT-Sicherheitsinfrastrukturen)

How to cite

APA:

Vaidya, A.M., Bergmann, S., & Riess, C. (2026). Bayesian Source Attribution of Synthetic Images. In Volker Skwarek, Kai Hendrik Wohnert, Sascha Kaven, Steffen Wendzel (Eds.), Lecture Notes in Informatics (LNI), Proceedings - Series of the Gesellschaft fur Informatik (GI) (pp. 141-152). Hamburg, DE: Gesellschaft fur Informatik (GI).

MLA:

Vaidya, Apoorva Mahipati, Sandra Bergmann, and Christian Riess. "Bayesian Source Attribution of Synthetic Images." Proceedings of the 13th Die Fachtagung Sicherheit, Schutz und Zuverlassigkeit, SICHERHEIT 2026, Hamburg Ed. Volker Skwarek, Kai Hendrik Wohnert, Sascha Kaven, Steffen Wendzel, Gesellschaft fur Informatik (GI), 2026. 141-152.

BibTeX: Download