M.A. Thesis:

Designing a Diagnosis System for Diabetes Using Significance-Based Feature Selection and Recurrent Neural Network Classification Algorithm

Today, the use of intrusion detection systems based on machine learning algorithms is one of the most accurate approaches to intrusion detection in the haze, one of the most important challenges in these systems is the increased accuracy of attack detection. In this study, a model is presented to solve this problem, where the trained intrusion detection system is deployed in the cloud and then receives the information received from the network layers, performs the control and detection process of possible attacks, as well as enhances security in the cloud. Sending information over the network and since the connection between the user layer nodes and the node nodes is one of the places most vulnerable to malicious attacks, this section uses an improved encryption algorithm to exchange information between the node nodes and the mobile wireless network, Algor The orphan used to encrypt the data here is an improved RSA algorithm. In fact, the proposed intrusion detection system consists of two parts, feature selection and classification, and the method is to first select the final features using a genetic-based feature selection method, and then to use a based classification model. Based on the Randomization Method, we perform the data classification process, after implementing and evaluating the proposed model, the overall accuracy value is 99.67%, the total sensitivity of the data is 99.77%, as well as the overall accuracy of the data is 99.85%Also in the network encryption section to improve the RSA algorithm in this Data encryption algorithm to process a method Fliplr () is used to change the ASCII code corresponding to the input message to the security of the data encryption algorithm to increase.

Keywords: Keywords: Parallel feature selection method, Fog-based intrusion detection system, Random assembly method