QR steganographic image transmission system based on multimode fiber and deep learning

Area of Science:

• Optical communication systems
• Digital image processing
• Steganography and cryptography

Background:

• Multimode optical fibers suffer from mode dispersion, causing image distortion during transmission.
• Existing steganographic methods lack robustness against transmission errors and security vulnerabilities.
• Deep learning offers potential solutions for image reconstruction and secure data embedding.

Purpose of the Study:

• To develop a novel QR steganographic image transmission system using deep learning over multimode optical fibers.
• To address and mitigate image distortion caused by mode dispersion.
• To enhance the security, reliability, and data recovery capabilities of image transmission.

Main Methods:

• Utilized the PIES-Net model for generating visually imperceptible steganographic images by embedding secret images into camouflaged images.
• Converted steganographic images into QR codes, leveraging their inherent error correction capabilities.
• Proposed an improved SFNet model based on U-Net architecture for reconstructing QR codes from scattered images at the receiving end.

Main Results:

• The developed system achieved high steganography with visually imperceptible embedded images.
• Recovered secret images demonstrated superior visual quality, peak signal-to-noise ratio, and image correlation.
• The system exhibited excellent robustness and security across various noise environments.

Conclusions:

• The proposed system effectively overcomes image distortion in multimode optical fiber transmission.
• Deep learning models (PIES-Net and SFNet) significantly improve the quality and security of steganographic image transmission.
• This research offers valuable contributions to multimode fiber communication and advanced image steganography techniques.