Research On Cloud Security Situation Awareness Method Based On Deep Learning

With the highly integrated development of deep learning,cloud computing and big data,the huge amount of data generated in the cloud computing application service layer,hundreds of millions of network access points and a variety of physical devices constitute a complex network space aggregation.The traditional security strategy becomes inefficient in the face of TB level data,relying on strong adaptability and self Deep learning of learning ability provides a new solution to cloud security problems.Traditional security technology protection measures are mainly passive defense against network attacks,which can not achieve a comprehensive real-time security analysis in the cloud situation awareness scene.This paper focuses on the establishment of a feasible security situation assessment and prediction model in the cloud computing environment to provide a basis for security situation deduction.In view of the complex data indicators in the cloud environment to achieve a reasonable classification,this paper selects vulnerability situation indicators represented by cloud security vulnerabilities and cloud system steady-state survivability situation indicators as situation indicator data,and applies the cloud security situation assessment method based on long-term and short-term memory network to the cloud system assessment,which can avoid the single-level traditional method situation assessment mode,from multiple From the perspective of service situation assessment and host situation assessment,the cloud system situation assessment is discussed in an all-round way.The expected alarm verification mechanism is added to the assessment,which can effectively reduce the invalid alarm information from the outside and reduce the impact on the later situation assessment.The cloud security situation data set is trained by longterm and short-term memory network,and the situation value is divided into appropriate cloud security state domain.Finally,compared with the traditional security situation level assessment method,the assessment results are more accurate.Aiming at the problem that the prediction accuracy of traditional network security situation prediction model is not high under the multi-source and heterogeneous big data environment,combined with the deep neural network in deep learning,this paper proposes a cloud security situation prediction model based on deep belief network.In this model,the mapping relationship between situation elements and prediction values is realized by deep belief network,and the parameters of hidden layer network are optimized by improved differential evolution algorithm.At the same time,two-dimensional rotation crossover strategy is introduced to increase the diversity of evolutionary population,and the convergence of the algorithm is verified to avoid premature convergence of situation prediction model.Finally,the simulation results show that the prediction accuracy is improved compared with the existing cloud security situation prediction model.