Abstract
With the enhancement of artificial intelligence and machine learning, nowadays it is in trend that mobile clients acquiesce their sign for diagnosing medical illnesses. Edge computing methodology is frequently used for medical diagnosis as it reduces the transmission latency and allows users and devices even at remote locations to analyze the data at the edge of the network. Since data-driven machine learning algorithms need to develop an identification system over huge medical data, they may be concerned about the privacy of data leakage during medical diagnosis. To solve this issue in our work a privacy-preserving medical diagnosis system is developed on edge. With our model, we encrypt the user input during the submission of the user input and that will be diagnosed for the disease and the user will be given the result in encrypted form which he could decrypt, to preserve the privacy of the user. The security and experimental analysis of our model explains the efficiency of our proposed system. The gradient boosting (CatBoost) model is redesigned by following the cloud-edge model, which accepts the ciphered model parameters rather than usual data to get rid of the amount of cipher to plain text computation using Triple-DES. In addition, we have optimized our model using the Harris Hawk Optimisation technique. Additionally, our algorithm offers private and prompt diagnosis while maintaining secure diagnosis on the edge. Our security study and investigational assessment depict that our algorithm is effective, efficient and secure.
Keywords: Catboost, Ciphertext, DES, Gradient boosting, Harris hawk optimization.