Research On Network Security Situation Assessment And Forecast Based On Multi-source Data Fusion

Cloud computing,Internet of things,the development of new technologies such as artificial intelligence,superposition and rapid evolution makes the network security risk fusion,resulting in the cyberspace threats are increasing,the network security situational awareness of multi-source data fusion technology can collect different safety equipment log or warning information,by analyzing the history of the security incident,Using situation assessment technology to get the current network security situation value to predict the future security trend,which is of great significance to the decision-making of network security administrators.In this paper,deep learning technology is used to conduct in-depth analysis and research on network security situation assessment and prediction methods.The main research contents are as follows:Firstly,aiming at the prominent problems of low precision and high false alarm rate in network attack detection of network security situation detection and evaluation analysis method,this paper mainly proposes a security situation evaluation method based on network attack detection.Firstly,the long-term and short-term memory recurrent neural network is used as the basic model(generator and discriminator)to capture the temporal correlation of time series distribution.At the same time,an outlier score is designed to detect the type of network attack through reconstruction error and discrimination error.Finally,according to the attack probability to determine the severity of the threat to the network,combined with the total loss caused by the network attack to quantify the network security situation.Secondly,according to the characteristics of network security situation prediction data timing,this paper selects long-term and short-term memory recurrent neural network to predict network security situation.In order to solve the problem of low accuracy of network security situation prediction,this paper introduces the weighted reinforcement mechanism in network security situation prediction when studying the structure of LSTM neural network.In this paper,we change the activation function on the basis of the original,introduce the sigmoid weighted linear unit to deal with the gradient problem in the back propagation,and multiply the input value by the sigmoid activation function,so that LSTM has a more complex structure to capture the recursive relationship between the input layer and the hidden layer.At the same time,aiming at the problem that the determination of the super parameters of LSTM neural network is mainly subjective selection,this paper introduces cuckoo search algorithm to automatically optimize the super parameters to improve the accuracy of the model.Thirdly,the CIC-IDS2017 data set published by Canadian research institute is used to verify the situation assessment and prediction method proposed in this paper.By building an intrusion detection environment to obtain network security events,the network security situation assessment method based on attack detection proposed in this paper is used to analyze the security situation when the network is attacked,and the situation value after situation assessment is further combined with the network security situation prediction method based on sigmoid weighted enhanced LSTM proposed in this paper to predict the future network security trend.The correctness and effectiveness of the model are verified through the comparative analysis of multiple sets of data.