URSMamba: Universal remote sensing image steganography using state space model

Area of Science:

• Computer Science
• Information Security
• Remote Sensing

Background:

• Image steganography is crucial for secure data transmission.
• Remote sensing images present unique challenges due to complex distributions and multiple spectral bands.
• Existing methods often overlook the specific requirements of remote sensing image steganography.

Purpose of the Study:

• To develop a universal steganography method for remote sensing images.
• To address the challenges of concealing and revealing abilities and hiding capacity in complex remote sensing data.
• To improve the performance of image steganography on multi-spectral remote sensing images.

Main Methods:

• Proposed URSMamba, a State Space Model-based universal remote sensing image steganography.
• Introduced Low-High Frequency Mamba Block (LHfreMB) for global and local feature extraction.
• Utilized Spectral Mamba Block (SpectralMB) for rich spectral information extraction.
• Developed Spatial-Spectral Dynamic Fusion (SSDF) block for adaptive feature integration.

Main Results:

• URSMamba demonstrates superior performance on multi-spectral remote sensing images.
• Achieved significant improvements in Peak Signal-to-Noise Ratio (PSNR) for cover/stego and secret/recovery image pairs.
• Outperformed state-of-the-art methods, showing 0.54 dB and 1.65 dB gains for 8-band images.
• Also delivered high-quality results on natural images with 3.25 dB and 2.96 dB improvements.

Conclusions:

• URSMamba effectively models complex ground distributions and spectral features in remote sensing images.
• The proposed method offers enhanced concealing and revealing capabilities for secure remote sensing data.
• URSMamba provides a robust and high-performance solution for remote sensing image steganography.