- 期刊
A Deep Learning Algorithm for Detecting Bot Infections in Host Networks
摘要
Botnets have become one of the main problems of current Internet security risks. Attackers can launch attacks based on botnets to induce online fraud, network information leakage, and other behaviors. Traditional botnet detection methods are not flexible, so this paper used the convolutional neural network (CNN), a deep learning algorithm, to detect bot program-infected host networks. Through data collection and pre-processing, a network model was built to perform feature learning of domain name bytes to detect whether the host network is infected with a botnet. The effectiveness of the CNN model in detecting botnets was verified by comparing it with the results of random forest and support vector machine. It was found that the CNN model had a precision of 97.14%, a recall rate of 97.43%, an F1 value of 97.28%, and an average recognition speed of 2.05 s, all of which were higher than those of both the random forest and support vector machine approaches. The results prove that the CNN model possesses high accuracy for detecting botnets and can be used for detecting bot programs in host networks.