基于改进LSTM方法的安全态势感知模型研究

Research on Security Situation Awareness Model Based on Improved LSTM Method

摘要: 网络环境中的各种网络攻击行为给网络带来了很多挑战，导致网络故障和负载增加等突发网络安全事件发生的概率变大，网络安全预警的前提是安全态势。因此，针对网络安全态势的不确定性、波动性等特点，提出了改进的长短期记忆(LSTM)网络的安全态势感知模型。首先，针对神经网络训练过程中速度较慢和数据维度过高的问题，采用卷积神经网络进行降维，然后利用改进的循环神经网络进行预测态势值，最后通过计算均方根误差来评价模型的优势。通过仿真对比实验验证了改进的LSTM模型大大降低了模型预测误差，能够更加高效、准确地实现对网络态势的评估和预测。

Abstract: Various network attack behaviors in the network environment have brought many challenges to the network, leading to increased probability of sudden network security incidents such as network failures and load increases. The prerequisite for network security early warning is the security situation. Therefore, in view of the uncertainty and volatility of the network security situation, an improved long short-term memory (LSTM) network security situation awareness model is proposed. First of all, in view of the slow speed and high dimensionality of the neural network training process, the convolutional neural network is used to reduce the dimension, then the improved recurrent neural network is used to predict the situation value, and finally the root mean square error is calculated to evaluate the model advantages. Simulation and comparison experiments verify that the improved LSTM model greatly reduces the model prediction error, and can more efficiently and accurately realize the evaluation and prediction of the network situation.