Deep Learning-Assisted Noise-Robust Optical Encryption Enabled by Pentachromatic Phototransistors

Optical communication requires effective encryption strategies for secure information transmission. The use of different wavelengths allows optical information to be configured in different ways, while broadband photoresponse enables these wavelength signals to be detected within a single device. Based on this approach, this study fabricated a broadband transparent phototransistor that operates across ultraviolet to infrared wavelengths. The IGZO/Ag2O/IGZO stacked structure exhibited a lower oxygen-related defect component and improved device stability. The responsivity increased by a factor of 2.38 and the noise equivalent power decreased by a factor of 1.95 compared to the IGZO/Ag2O device. The broadband photoresponse properties of the device enabled the separation of key and data channels, while a dynamic key enhanced security. This approach restricts pattern recognition compared to using fixed keys, resulting in a higher security rate. The learning-based approach was evaluated under noise and sensitivity-drift conditions, together with the effect of photoresponse stability under repetitive operation. These results contribute to the fabrication of broadband transparent phototransistors, facilitating high-performance optical encryption systems.