The Cover Source Mismatch Problem in Deep-Learning Steganalysis

Giboulot Quentin, Bas Patrick, Cogranne Rémi, Borghys Dirk

Résultats de recherche: Chapitre dans un livre, un rapport, des actes de conférencesContribution à une conférenceRevue par des pairs

Résumé

This paper studies the problem of Cover Source Mismatch (CSM) in steganalysis, i.e. the impact of a testing set which does not originate from the same source than the training set. In this study, the trained steganalyzer uses state of the art deep-learning architecture prone to better generalization than feature-based steganalysis. Different sources such as the sensor model, the ISO sensitivity, the processing pipeline and the content, are investigated. Our conclusions are that, on one hand, deep learning steganalysis is still very sensitive to the CSM, on the other hand, the holistic strategy leverages the good generalization properties of deep learning to reduce the CSM with a relatively small number of training samples.

langue originaleAnglais
titre30th European Signal Processing Conference, EUSIPCO 2022 - Proceedings
EditeurEuropean Signal Processing Conference, EUSIPCO
Pages1032-1036
Nombre de pages5
ISBN (Electronique)9789082797091
étatPublié - 2022
Evénement30th European Signal Processing Conference, EUSIPCO 2022 - Belgrade, Serbie
Durée: 29 août 20222 sept. 2022

Série de publications

NomEuropean Signal Processing Conference
Volume2022-August
ISSN (imprimé)2219-5491

Une conférence

Une conférence30th European Signal Processing Conference, EUSIPCO 2022
Pays/TerritoireSerbie
La villeBelgrade
période29/08/222/09/22

Empreinte digitale

Examiner les sujets de recherche de « The Cover Source Mismatch Problem in Deep-Learning Steganalysis ». Ensemble, ils forment une empreinte digitale unique.

Contient cette citation