Abstract:Image steganography is the process of hiding secret data in a cover image by subtle perturbation. Recent studies show that it is feasible to use a fixed neural network for data embedding and extraction. Such Fixed Neural Network Steganography (FNNS) demonstrates favorable performance without the need for training networks, making it more practical for real-world applications. However, the stego-images generated by the existing FNNS methods exhibit high distortion, which is prone to be detected by steganalysis tools. To deal with this issue, we propose a Cover-separable Fixed Neural Network Steganography, namely Cs-FNNS. In Cs-FNNS, we propose a Steganographic Perturbation Search (SPS) algorithm to directly encode the secret data into an imperceptible perturbation, which is combined with an AI-generated cover image for transmission. Through accessing the same deep generative models, the receiver could reproduce the cover image using a pre-agreed key, to separate the perturbation in the stego-image for data decoding. such an encoding/decoding strategy focuses on the secret data and eliminates the disturbance of the cover images, hence achieving a better performance. We apply our Cs-FNNS to the steganographic field that hiding secret images within cover images. Through comprehensive experiments, we demonstrate the superior performance of the proposed method in terms of visual quality and undetectability. Moreover, we show the flexibility of our Cs-FNNS in terms of hiding multiple secret images for different receivers.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that the steganographic images generated by the existing Fixed Neural Network Steganography (FNNS) methods have high distortion and are easily detected by steganalysis tools. To solve this problem, the author proposes Cover - Separation Fixed Neural Network Steganography (Cs - FNNS) based on the deep generative model. Specifically, Cs - FNNS improves the existing FNNS methods in the following ways: 1. **Introducing AI - generated cover images**: Use the deep generative model to generate cover images. These images will not arouse suspicion when transmitted and shared on the network, thereby improving the concealment of steganography. 2. **Proposing the Steganographic Perturbation Search (SPS) algorithm**: This algorithm directly encodes the secret data into imperceptible perturbations and combines them with the AI - generated cover images to generate steganographic images. The goal of the SPS algorithm is to find a perturbation so that the steganographic image is visually close to the cover image, can trigger a fixed random decoding network to output the secret data at the same time, and can deceive steganalysis tools. 3. **Separating the decoding process**: Different from the traditional FNNS method that directly extracts secret data from the steganographic image, Cs - FNNS first regenerates the cover image using the shared key at the receiving end, then separates the perturbation from the steganographic image, and then inputs the perturbation into the decoding network to recover the secret data. This strategy eliminates the interference of the cover image on the decoding network and encourages Cs - FNNS to search for smaller perturbations, thereby reducing the distortion of the steganographic image. Through these improvements, Cs - FNNS performs excellently in terms of visual quality and undetectability, and has high flexibility. It can hide multiple secret images in a single cover image for different receivers to decode. Experimental results show that Cs - FNNS is significantly superior to the existing FNNS methods in terms of visual quality and secret image recovery.

Cover-separable Fixed Neural Network Steganography via Deep Generative Models

Securing Fixed Neural Network Steganography

A Novel Grayscale Image Steganography via Generative Adversarial Network

Generative adversarial networks-based image steganography with multiscale features integration

Batch Steganography Via Generative Network

Steganography of Steganographic Networks

Generative Steganography Network

Large-capacity and Flexible Video Steganography via Invertible Neural Network

Lossless Image Steganography Based on Invertible Neural Networks

Channel Attention Image Steganography With Generative Adversarial Networks

Deeply‐Recursive Attention Network for Video Steganography

Image steganography based on generative implicit neural representation

Lossless image steganography: Regard steganography as super-resolution

Spatial Image Steganography Based on Generative Adversarial Network

Preemptive Image Protection against Steganography

Cover Reproducible Steganography Via Deep Generative Models.

Coverless Information Hiding Based on Generative adversarial networks

Image Generation Network for Covert Transmission in Online Social Network

Hiding Data in Colors: Secure and Lossless Deep Image Steganography via Conditional Invertible Neural Networks

High invisibility image steganography with wavelet transform and generative adversarial network

A General Steganographic Framework for Neural Network Models.