Dynamic Image Encryption using Neural Networks for Medical Images
Adithya Krishna, Vanya Arikutharam, K Venkat Ramnan, H Bharathi, T. S. Chandar · 2022 IEEE IAS Global Conference on Emerging Technologies (GlobConET) · 2022
There is an increasing need to secure medical images for safe storage and transmission under the booming field of E-Healthcare and telemedicine. Medical images require utmost privacy due to the high level of sensitive information they carry. Most hospitals use AES - 256 to encrypt data in the current scenario. The key used for encryption is randomly generated, and at times this can pose problems to the strength of the encryption. This work proposes a novel cipher that strengthens the encryption using a dynamic key generated by a neural network for medical images. A novel encryption method in the lines of symmetric block encryption is proposed where the region of interest (ROI) of the given medical image is extracted and used as an input seed for a pseudo-random number generator (PRNG) generated using a generative adversarial network (GAN) and served as the key to the novel encryption algorithm. The National Institute of Standards and Technology (NIST) suite evaluates the GAN-based PRNG and the encryption results are benchmarked against existing fast cipher standards and evaluated with common methods like histogram analysis, NPCR values, information entropy, visual testing and complexity analysis.