摘要
为了提高网络安全态势预测准确率及稳定性,提出了基于深度信念网络的网络安全态势预测方法。网络安全态势预测包括两个阶段,首先通过逐层训练的方式预训练各层中受限玻尔兹曼机,经多层受限玻尔兹曼机之间的映射,完成权重和偏置参数更新,采用深度信念网络实现更新后参数的微调,然后利用训练完成深度信念网络模型,将网络攻击行为作为深度信念网络的输出向量,完成网络安全势态预测,最后采用具体网络入侵数据进行仿真测试。测试结果表明,深度信念网络的网络安全态势预测精度高,而且预测结果十分稳定。
In order to improve the accuracy and stability of network security situation prediction,a network security situation prediction method based on deep belief network is proposed.Network security situation prediction consists of two stages.Firstly,the restricted Boltzmann machines in each layer are pre trained by layer by layer training.Through the mapping between multi-layer restricted Boltzmann machines,the weight and bias parameters are updated.Then,the parameters are fine tuned by using the deep belief network.Then,the deep belief network model is completed by training,and the network attack behavior is regarded as the deep belief network.Finally,the specific network intrusion data is used for simulation test.The test results show that the network security situation prediction accuracy of deep belief network is high,and the prediction result is very stable.
作者
刘东伟
Liu Dongwei(Hebei Zhongyan Industry Co.,Ltd.,Shijiazhuang 050051,China)
出处
《国外电子测量技术》
2020年第12期44-48,共5页
Foreign Electronic Measurement Technology
关键词
深度信念网络
网络安全
态势预测
训练样本
数据集
学习表示
参数更新
deep belief network
network security
situation prediction
training samples
data set
learning representation
parameters are updated