Encryption and decryption applications on conductive films using THz real-time high-resolution imaging.

Area of Science:

• Optics and Photonics
• Information Security
• Artificial Intelligence

Background:

• Terahertz (THz) imaging offers unique properties for secure applications.
• Material-based encryption using THz transmission is feasible but challenged by low signal-to-noise ratio (SNR) imaging.
• Existing denoising methods struggle with THz imaging clarity degradation.

Purpose of the Study:

• To develop a real-time THz quasi-near-field imaging system for low-cost, concealable information encryption and decryption.
• To address the challenge of imaging clarity deterioration under ultra-low SNR conditions.
• To enhance the stability and recognizability of encrypted information in noisy environments.

Main Methods:

• Designed three types of encryption structures (text, coating, QR code) using graphite pencils with distinct THz transmission properties.
• Proposed a modified U-Net deep learning denoising method incorporating multi-frequency feature fusion and THz-specific physical priors.
• Evaluated performance against the block-matching and 3D filtering (BM3D) algorithm using metrics like PSNR, SSIM, and FSIM.

Main Results:

• Successfully decrypted all designed encryption structures, validating material-based THz encryption.
• The modified U-Net method significantly suppressed noise while preserving fine structural details, outperforming BM3D.
• Encrypted information remained clearly recognizable even in complex noise environments, showing improved PSNR, SSIM, and FSIM values.

Conclusions:

• A unified framework combining material-based encryption, THz imaging, and deep learning denoising was established.
• The proposed approach offers low implementation cost, strong concealment, and enhanced stability for THz imaging applications.
• This work paves the way for secure communication, anti-counterfeiting, and information protection using THz technology.