Optimization of Intrusion Detection System Based on Improved Convolutional Neural Network Algorithm

Abstract

The commonly used method in network intrusion detection is abnormal behavior detection, but because abnormal behavior detection is based on artificially setting abnormal values for judgment, the efficiency is low and the false alarm rate is high. For this problem, an intrusion detection system architecture combining K-means algorithm and convolutional neural network algorithm has been introduced. First, the data stream is clustered through the K-means algorithm, and the abnormal data is initially separated, and then the data is applied to the convolution. In the neural network algorithm, the intrusion data flow is judged. The experimental results prove that the framework improves the efficiency of the intrusion detection system to a certain extent and effectively improves the detection accuracy.