A Deep Learning-Based Stream Cipher Generator for Medical Image Encryption and Decryption

In particular, the necessity for medical image encryption is becoming more and more important to protect patient privacy about their medical imaging data. The private key generated by this unique deep learning-based key generation network (DeepKeyGen) stream cipher generator might be used to encrypt and decrypt medical images. The sort of learning network used in DeepKeyGen to produce the private key is the generative adversarial network (GAN). Moreover, the learning network is intended to guide realizing the private key generation process by the transformation domain, which stands for the "style" of the private key to be generated. DeepKeyGen aims to locate the mapping link between the original image and the private key. DeepKeyGen is evaluated using three distinct data sets: the Montgomery Township chest X-ray, the BraTS18, and the Ultrasonic Brachial Plexus. The evaluation's findings and security analysis show that the recommended key generation circuit can generate a key that is private with a high level of security. This paper provides a novel approach to encrypt and decrypt images from hospitals using a Deep Learning-Based Stream Cypher Ger (DL-SCG). The DL-SCG strengthens the security of medical image transmission by utilizing deep neural network capabilities to produce cryptographic keys in the form of a stream cipher dynamically. To further strengthen the overall security of the encryption process, the DL-SCG's architecture is painstakingly designed to autonomously learn and generate secure and unexpected key streams. The DLSCG adapts to the unique properties of medical images, such as different resolutions, while simultaneously guaranteeing a high degree of security through the integration of convolutional and recurrent neural networks.