SteganoCNN: Image Steganography with Generalization Ability Based on Convolutional Neural Network

Area of Science:

• Computer Science
• Artificial Intelligence
• Image Processing

Background:

• Image steganography involves concealing one image within another.
• Embedding multiple secret images into a single carrier image presents a significant challenge.
• Deep learning applications in practical image steganography are currently limited.

Purpose of the Study:

• To propose a novel deep learning model, SteganoCNN, for embedding two secret images into one carrier image.
• To enable effective reconstruction of both secret images from the stego-image.
• To address the limitations of current image steganography techniques.

Main Methods:

• Development of a Steganography Convolution Neural Network (SteganoCNN) comprising an encoding and a decoding network.
• The decoding network incorporates two distinct extraction networks.
• End-to-end training of the entire network for simultaneous embedding and extraction.
• The encoding network embeds secret images, while the decoding network reconstructs them.

Main Results:

• The proposed SteganoCNN model achieves a maximum payload capacity of 47.92 bits per pixel.
• The steganography scheme effectively evades detection by steganalysis tools.
• The stego-image quality remains undistorted after the embedding process.
• SteganoCNN demonstrates strong generalization capabilities across different image types, including remote sensing and aerial images.

Conclusions:

• SteganoCNN offers a robust solution for embedding and reconstructing two secret images within a single carrier image.
• The model provides high security and payload capacity, making it suitable for real-world applications.
• Its ability to handle diverse image data types enhances its practical utility in advanced image steganography.