基于卷积神经网络与跳跃连接的网络入侵检测系统

Network Intrusion Detection System Based on Convolutional Neural Network and Skip Connections

网络入侵检测系统(NIDS)在保护计算机网络中扮演着至关重要的角色。现有的方法不能持续性地检测新型攻击行为。此外,手工设计特征提取是繁琐的并且无法选择出非常适合的特征进行网络入侵检测。为了解决上述挑战,提出一种新颖的基于卷积神经网络的入侵检测模型。该方法能自动化地提取复杂高维的特征,并且引入跳跃链接克服神经网络训练的过拟合问题,从而实现高准确率。实验显示,提出的方法在KDD99数据集下取得98.33%的准确率,优于基于传统的机器学习方法。

Network Intrusion Detection System(NIDS)plays a vital role in protecting computer networks.Existing methods cannot continuously detect new types of attacks.In addition,manually designing feature extraction is tedious and cannot select very suitable features for network intru⁃sion detection.In order to solve the above challenges,a novel intrusion detection model based on convolutional neural networks is pro⁃posed.This method can automatically extract complex high-dimensional features,and introduce jump links to overcome the over-fitting problem of neural network training,thereby achieving high accuracy.Experiments show that the proposed method achieves an accuracy of 98.33%on the KDD99 data set,which is superior to traditional machine learning methods.