In recent years, optical computing has been gaining more attention due to its potential advantages in speed, energy efficiency, and scalability. Among various methods of optical computing, diffractive deep neural networks (D2NNs) have brought forward as an emerging free-space platform for optical computing. D2NNs use deep learning methods to design spatially structured diffractive surfaces that can modulate the light diffraction to compute a given task at the speed of light propagation.
Apart from its speed and energy efficiency, D2NNs also provide unique advantages for visual computing tasks, as they are capable of directly processing and accessing 2D and 3D spatial information of a scene. This direct access to optical information makes them ideal for visual computing tasks such as image classification, hologram reconstruction, quantitative phase imaging, and seeing through random diffusers.
A group of researchers at the University of California, Los Angeles (UCLA), recently presented a diffractive network that performs data class-specific transformations and optical image encryption. The diffractive networks were trained using deep learning, and after their training was completed, they were physically fabricated using 3D printing to optically transform the input images and produce encrypted, uninterpretable output patterns captured by an image sensor. Only by applying the correct decryption keys can the encrypted images be restored to reveal the original information. Applying mismatched inverse transformations using incorrect decryption keys results in various noise-like patterns.
The researchers demonstrated that this class-specific all-optical image encryption design is feasible at both near-infrared and terahertz wavelengths, making it possible to operate at various illumination wavelengths. By utilizing a class-specific transformation matrix, the encryption scheme provides an additional layer of security, making it more difficult to decipher the original images that belong to the target data class by reverse engineering.
This class-specific design enables secure data distribution to multiple end-users using only one diffractive encryption network. By distributing different decryption keys to various recipients based on their data access permissions, it ensures that only the desired portion of the input data is shared with authorized users, even though a single diffractive network optically encrypts all different data classes. This feature ensures secure data distribution, while the system is attractive for distributed image encryption with a fast, task-specific, and energy-efficient all-optical encryptor.
This new diffractive image encryption design provides further opportunities for developing image and data security solutions and all-optical image processing devices that work at various illumination wavelengths. However, this technology may still have some issues, such as scalability and flexibility. Further research might be necessary to address these challenges and push this technology further.
In conclusion, the diffractive image encryption design developed by the UCLA research team could have significant implications for various areas, including national security, communications, and finance. It provides a secure, efficient, and energy-saving way of encrypting and distributing sensitive data while keeping it confidential and secure from unauthorized access. The application of this technology may be unlimited, depending on the imagination and creativity of researchers in the respective fields.
<< photo by Tima Miroshnichenko >>
