An Efficient Encryption and Compression of Sensed IoT Medical Images Using Auto-Encoder

Healthcare systems nowadays depend on IoT sensors for sending data over the internet as a common practice. Encryption of medical images is very important to secure patient information. Encrypting these consumes lot time edge computing; therefore, use an auto-encoder compression before encoding will solve such problem. In this paper, we compress image encryption, and encryption output (vector) sent out network. On other hand, decoder was used reproduce original back after vector received decrypted. Two convolutional neural networks were conducted evaluate our proposed approach: The first one auto-encoder, which utilized encrypt images, assesses classification accuracy decryption decoding. Different hyperparameters encoder tested, followed by verify that no critical information lost, test resolution. approach, sixteen hyperparameter permutations are utilized, but research discusses three main cases in detail. case shows combination Mean Square Logarithmic Error (MSLE), ADAgrad, two layers ReLU had best results with Absolute (MAE) = 0.221 50 epochs 75% result algorithm. second reflection (MSE), RMSprop, ReLU, 65%, gives MAE 0.31 epochs. third worst, hinge, providing 20% 0.485.