基于深度置信网络的网络安全态势感知与预测

Network Security Situation Awareness and Prediction Based on Deep Belief Network

针对网络安全防御要素和评估结果的不确定性和模糊性问题,提出一种基于深度置信网络的特征提取方法,通过多层玻尔兹曼机进行特征降维并提取影响网络安全的关键因素,然后在深度置信网络最后一层加上softmax类别分类器,实现对网络安全态势感知和预测。真实的网络环境实验表明,基于深度置信网络的非监督方法能够在一定程度上提高网络安全的预测精度,提高模型的可用性和有效性。

In view of the uncertainty and fuzziness of network security defense elements and evaluation results, a feature extraction method based on deep belief network is proposed. Through hierarchical Boltzmann machine, feature dimensionality reduction and key factors affecting network security are extracted. Then, a soft Max classifier is added to the last layer of deep belief network to realize network security situational awareness and prediction. The real network environment experiments show that the unsupervised method based on deep confidence network can improve the prediction accuracy of network security to a certain extent and improve the availability and effectiveness of the model.