EVALUATING THE SECURITY PERFORMANCE OF USING NEURAL NETWORK IN INDUSTRIAL INTERNET OF THINGS

Objective:  The aim of this research is to develop group-based classifications that improve the accuracy of intrusion detection. Method: So far, various methods and algorithms have been presented for data mining, and in this research, machine learning, neural network and optimization of weights with the help of genetic algorithm were used. For this purpose, kdd 99 dataset was used for data preprocessing and Convolutional Neural Network (CNN) algorithm was evaluated along with SVM and KNN algorithms. SVM and KNN algorithms were implemented using R software and convolutional neural network algorithm in MATLAB. Results: The results show that the use of machine learning and the neural system have acceptable results in intrusion detection. However, the weights optimized by the genetic algorithm show higher accuracy than the case with the original weights. Novelty: Comparing these results with the results of previous studies shows that the speed of convergence is much faster, which is one of the advantages of this algorithm.