A study on the invariance in security whatever the dimension of images for the steganalysis by deep-learning - LIRMM - Laboratoire d’Informatique, de Robotique et de Microélectronique de Montpellier
Preprints, Working Papers, ... Year : 2023

A study on the invariance in security whatever the dimension of images for the steganalysis by deep-learning

Marc Chaumont
Connectez-vous pour contacter l'auteur
Frédéric Comby

Abstract

In this paper, we study the performance invariance of convolutional neural networks when confronted with variable image sizes in the context of a more "wild steganalysis". First, we propose two algorithms and definitions for a fine experimental protocol with datasets owning "similar difficulty" and "similar security". The "smart crop 2" algorithm allows the introduction of the Nearly Nested Image Datasets (NNID) that ensure "a similar difficulty" between various datasets, and a dichotomous research algorithm allows a "similar security". Second, we show that invariance does not exist in state-ofthe-art architectures. We also exhibit a difference in behavior depending on whether we test on images larger or smaller than the training images. Finally, based on the experiments, we propose to use the dilated convolution which leads to an improvement of a state-of-the-art architecture.
Fichier principal
Vignette du fichier
ICASSP2023_PLANOLLES_CHAUMONT_COMBY_Security_Invariance.pdf (765.16 Ko) Télécharger le fichier
QA-ICASSP2023.pdf (190.75 Ko) Télécharger le fichier
Origin Files produced by the author(s)
Licence
Public Domain
Licence
Public Domain

Dates and versions

lirmm-04001355 , version 1 (23-02-2023)

Licence

Public Domain

Identifiers

Cite

Kévin Planolles, Marc Chaumont, Frédéric Comby. A study on the invariance in security whatever the dimension of images for the steganalysis by deep-learning. 2023. ⟨lirmm-04001355⟩
34 View
56 Download

Altmetric

Share

More