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

TL;DR

This study investigates whether convolutional neural networks maintain consistent performance across different image sizes in steganalysis, revealing a lack of invariance and proposing dilated convolutions for improvement.

Contribution

The paper introduces algorithms and datasets for consistent difficulty and security in steganalysis, and demonstrates the benefits of dilated convolutions in neural network architectures.

Findings

Invariance does not exist in current architectures.

Performance varies with image size relative to training images.

Dilated convolutions improve steganalysis accuracy.

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-of-the-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.