Weakening the Detecting Capability of CNN-based Steganalysis

TL;DR

This paper introduces a method to generate adversarial examples that weaken CNN-based steganalysis, highlighting vulnerabilities and aiming to improve the security of steganographic detection methods.

Contribution

It presents a novel approach to create adversarial examples specifically for steganalysis CNNs, demonstrating a new way to challenge existing detection algorithms.

Findings

Adversarial examples significantly increase detection error rates.

Proposed method effectively reduces CNN steganalysis accuracy.

Experiments confirm the method's ability to weaken detection capability.

Abstract

Recently, the application of deep learning in steganalysis has drawn many researchers' attention. Most of the proposed steganalytic deep learning models are derived from neural networks applied in computer vision. These kinds of neural networks have distinguished performance. However, all these kinds of back-propagation based neural networks may be cheated by forging input named the adversarial example. In this paper we propose a method to generate steganographic adversarial example in order to enhance the steganographic security of existing algorithms. These adversarial examples can increase the detection error of steganalytic CNN. The experiments prove the effectiveness of the proposed method.