Protection Against Criminal use of Steganography – PACeS
Protection Again Criminal uses of Steganography
Detection of hidden information in images and images generated using artificial intelligence.
Protection Again Criminal uses of Steganography
- Detect hidden information in images<br />- Understand the factors that make this detection difficult<br />- Characterize the signature of a natural photograph<br />- Use the proposed method, based on the signature of a photograph's acquisition chain, to identify images generated by artificial intelligence
We characterized the signature of the acquisition chain of a natural photograph based on the statistical distribution of noise in residual images.
- We characterized the correlation between neighboring pixels as a signature of the acquisition chain of a natural photograph.
- We exploited the correlation structure between pixels to measure the difficulty of detecting hidden information.
- We exploited the correlation structure between pixels to explain the source mismatch (distribution shift) problem in steganography.
- We exploited the correlation structure between pixels to identify images generated by artificial intelligence.
Work to characterize the correlation between pixels and the chain of acquisition of a natural photograph in order to detect images generated by artificial intelligence is still in its infancy. We will continue in this direction with a possible application to videos.
- Antoine Mallet , Martin Beneš , Rémi Cogranne, «Cover-source mismatch in steganalysis: systematic review«, EURASIP Journal on Information Security, 2024, 2024 (1), pp.26. ?10.1186/s13635-024-00171-6?
- Antoine Mallet , Patrick Bas , Rémi Cogranne, «Statistical Correlation as a Forensic Feature to Mitigate the Cover-Source Mismatch«, 12th ACM Workshop on Information Hiding and Multimedia Security (ACM IH\&MMSEC'24), Jun 2024, Baiona, Spain. ?10.1145/3658664.3659638?
- Rémi Cogranne, «A Comparative Review of Deep Learning Models for Deepfake Detection«, 2025 IEEE International Conference on Advanced Machine Learning and Data Science (IEEE AMLDS 2025), IEEE, Jul 2025, Tokyo (Japan), Japan
- Arthur Méreur , Antoine Mallet , Rémi Cogranne , Minoru Kuribayashi, «FORENSICS ANALYSIS OF RESIDUAL NOISE TEXTURE IN DIGITAL IMAGES FOR DETECTION OF DEEPFAKE«, International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Apr 2025, Hyderabad, India, France
- Antoine Mallet , Rémi Cogranne , Minoru Kuribayashi , Arthur Méreur, «How much is the Source Mismatch an Important Problem for Deepfake Detection ?«, APSIPA Transactions on Signal and Information Processing
- Arthur Méreur , Antoine Mallet , Rémi Cogranne, «Are Deepfakes a Game Changer in Digital Images Steganography Leveraging the Cover-Source-Mismatch?«, The 19th International Conference on Availability, Reliability and Security, Jul 2024, Vienne (AUT), Austria. ?10.1145/3664476.3670893?
- Antoine Mallet , Arthur Méreur , Minoru Kuribayashi , Rémi Cogranne , Patrick Bas, «Simple Detection of AI-Generated Images based on Noise Correlation«, International conference on Advanced Machine Learning and Data Science (AMLDS) 2025, Jul 2025, Tokyo, Japan
- Quentin Giboulot , Patrick Bas , Rémi Cogranne , Dirk Borghys, «The Cover Source Mismatch Problem in Deep-Learning Steganalysis«, European Signal Processing Conference, Aug 2022, Belgrade, Serbia
- Rémi Cogranne, «A Comparative Review of Deep-Learning Models for Deepfakes Detection«, SPIE 10th International Conference on Multimedia and Image Processing (ICMIP 2025), Apr 2025, Okinawa, Japan, Japan
- Quentin Giboulot , Patrick Bas , Rémi Cogranne, «Multivariate Side-Informed Gaussian Embedding Minimizing Statistical Detectability«, IEEE Transactions on Information Forensics and Security, 2022, 17, pp.1841 - 1854. ?10.1109/TIFS.2022.3173184?
- Antoine Mallet , Rémi Cogranne , Patrick Bas , Quentin Giboulot, «Identification de Développements d'Images par Matrices de Corrélations«, XXIXème Colloque Francophone de Traitement du Signal et des Images, Université de Grenoble; Association Gretsi, Aug 2023, Grenoble, France
- Antoine Mallet , Rémi Cogranne , Patrick Bas, «Linking Intrinsic Difficulty and Regret to Properties of Multivariate Gaussians in Image Steganalysis«, 12th ACM Workshop on Information Hiding and Multimedia Security (ACM IH\&MMSEC'24), Jun 2024, Baiona, Spain. ?10.1145/3658664.3659643?
The present PACeS project focuses on steganography and steganalysis, i.e. data hiding within innocuous-like digital media. Such tools are readily available on the Internet and can be used by criminals rings, or for sensitive data leakage. Surprisingly, there is no software to detect this threat. This project aims essentially at developing such detection tools for filling this gap. Our method relies on accurate media statistical models for guaranteeing a very low false-positive rate and leveraging the design of adversarial methods, exploiting such models by hiding data while minimizing the detectability. PACeS project also addresses very operational issues by focusing on a context that is as realistic as possible. This includes, especially, addressing the problems related to the very high diversity of media as well as taking into account the increasing use of videos can be studied in steganography and steganalysis due to complete absence of exploitable statistical models.
Project coordination
Rémi COGRANNE (Laboratoire Informatique et Société Numérique Unité de recherche)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
Partnership
LIST3N Laboratoire Informatique et Société Numérique Unité de recherche
Thales / Thales Service -Etudes Amonts SiX
CRIStAL Centre de Recherche en Informatique, Signal et Automatique de Lille
Montimage (MTI) / R&D
Help of the ANR 194,351 euros
Beginning and duration of the scientific project:
- 42 Months
