Experimental optical encryption based on random mask encoding and deep learning.

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Information Security

Background:

• Traditional optical encryption methods often require complex optical setups or precise characterization of encryption elements.
• Deep learning offers a powerful approach for learning complex, non-linear mappings essential for advanced encryption.

Purpose of the Study:

• To develop and experimentally validate a new optical encryption scheme leveraging random mask encoding and deep learning.
• To demonstrate the feasibility of using a neural network to learn the encryption-decryption process without explicit filter characterization.

Main Methods:

• An optical encryption setup was designed using a random amplitude modulation with a binary mask fabricated on a polyethylene terephthalate film.
• A deep neural network was trained using 2500 object-speckle image pairs to learn the mapping between the original phase image and its encrypted speckle pattern.
• The trained neural network was employed for the decryption process to recover the original image from the ciphertext.

Main Results:

• The experimental results confirmed the successful implementation of the proposed optical encryption scheme.
• The deep learning-based decryption successfully reconstructed the primary image from the speckle pattern ciphertext with high fidelity.
• The scheme demonstrated efficient processing, quickly outputting the original image.

Conclusions:

• The proposed scheme effectively combines random mask encoding with deep learning for robust optical image encryption.
• This approach offers a practical and efficient method for optical encryption, reducing the need for explicit physical parameterization of the encryption mask.
• The successful demonstration paves the way for advanced, AI-driven optical security solutions.