摘要
入侵检测是检测和预防可能对基于网络的计算机系统进行攻击和入侵作出反应的技术。提出一种基于深度卷积神经网络的入侵检测的算法,在卷积神经网络基础上引入Inception模型和残差网络,采用深度学习技术,如Relu、Dropout、Softmax。提高模型的收敛速度,使得训练的模型的泛化能力更强,增加网络的宽度和深度,提升网络对尺度的适应性。使用KDD Cup 99数据对该算法进行验证,实验表明,该网络模型与GoogleNet和Lenet-5相比具有更高的准确率和检测率,准确率能够达到94.37%,误报率仅2.14%,提高了入侵检测识别的分类准确性。
Intrusion detection is the technology of detecting,preventing,and possibly reacting to attacks and intrusions in network-based computer systems.This paper proposes an intrusion detection algorithm based on deep convolutional neural network.It introduced Inception model and residual network based on convolutional neural network,and adopted deep learning techniques such as Relu,Dropout,and softmaxSoftmax.The convergence speed of the model was improved,the generalization ability of the trained model was stronger,the width and depth of the network were increased,and the adaptability of the network to the scale was improved.Using KDD Cup 99 data to verify the algorithm,the experiments show that compared with GoogleNet and Lenet-5,the network model has higher accuracy and detection rate,with an accuracy rate of 94.37%and a false alarm rate of 2.14%,which improves the classification accuracy of intrusion detection and recognition.
作者
李勇
张波
Li Yong;Zhang Bo(School of Automation,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)
出处
《计算机应用与软件》
北大核心
2020年第4期324-328,共5页
Computer Applications and Software
