The openness of network data makes it vulnerable to hackers, viruses and other attacks, which seriously threatens the privacy and property security of users. In order to improve the accuracy of the intrusion detection for network security communication, based on the traditional intrusion detection system, combining with the deep learning theory and shortcomings, this paper proposed an intrusion detection system for network security communication based on multi-scale convolutional neural network, and conducted the corresponding experiments on public datasets. The experimental results perform that compared to the intrusion detection system based on Adaboost model and Recurrent Neural Network model, the convergence speed of multi-scale convolutional neural network system is faster, the average error detection rate is reduced by 4.02%, and the average accuracy is improved by 4.37%. The results prove that the intrusion detection system based on multi-scale convolution neural network has a high detection accuracy.

网络数据的开放性使其容易受到黑客、病毒等攻击，严重威胁用户的隐私和财产安全。为了提高网络安全通信入侵检测的准确性，在传统入侵检测系统的基础上，结合深度学习理论和不足，提出了一种基于多尺度卷积神经网络的网络安全通信入侵检测系统。网络，并在公共数据集上进行了相应的实验。实验结果表明，与基于Adaboost模型和循环神经网络模型的入侵检测系统相比，多尺度卷积神经网络系统的收敛速度更快，平均错误检测率降低4.02%，平均准确率提高了4.37%。结果证明基于多尺度卷积神经网络的入侵检测系统具有较高的检测精度。