Improved BP Algorithm Based Intrusion Detection Method Research In The Cloud Environment

As a key technology of cloud computing, virtualization technology enables the efficient use of cloud computing resources.Due to the openness of cloud computing, the virtualization environment is facing some security threats. The detection efficiency of traditional intrusion detection technology is low and it is difficult to compatible with the network structure of the virtualized environments.Therefore, how to implement the intrusion detection in virtualized environments has important significance for the development of cloud computing virtualization technology.The existing cloud intrusion detection systems can not be applied to a virtualized environment that has different network models, and the detection accuracy of variant of attack is lower. We conducted in-depth research on intrusion detection method of cloud environments,and proposed two improved BP detection algorithms based on swarm intelligence algorithms(PSO and DE). In this research works, the following contributions were made:(1) Conducted in-depth study on intrusion detection methods in virtualized environment, summarized the technical characteristics of all kinds of methods, analysed the problems of existing intrusion detection techniques and methods of cloud.(2) To address the issues of existing methods, on the basis of in-depth study of the KVM network models, an improved BP algorithm based cloud environment intrusion detection model (MBPCIDM) was proposed. The model includes data capture module, packet parsing module, feature extraction module, neural network detection engine, intrusion response module etc. It is compatible with KVM virtualization environment that has different network models,and is able to provide intrusion detection services for cloud environments.(3) Based on in-depth analysis of virtualized network structure, due to the problem of BP algorithm that it is easy to fall into local minima, an PSO algorithm based improved BP detection algorithm is proposed. The algorithm combines the ability of searching global optimal solution of PSO and the feature of the gradient descent in local search of BP algorithm, and the PSO algorithm was introduced to optimize of the initial weight and threshold of BP, the use of adaptive learning rate and momentum factor method could accelerate the convergence speed of BP neural network and prevent it from falling into local minimum. Experimental results show that the average detection rate of the intrusion detection algorithm constructed by the improved algorithm is higher, can provide intrusion detection services for cloud computing environments effectively and reliably.(4) To deal with the premature convergence problem of PSO algorithm, make use of differential evolution algorithm(DE) to improve the BP algorithm, a DE algorithm based BP detection algorithm was proposed. The detection algorithm use DE to search the optimal initial weights and thresholds of neural network firstly, then use the BP algorithm to fine-tune the weights. The weights and thresholds were updated in each generation to minimize the mean square error(MSE). The experimental results show that under appropriate parameter setting conditions, DE algorithm is superior to PSO algorithm on optimizing the initial weights and thresholds of BP neural network.The study in the thesis is an useful attempt for the intrusion detection techniques and methods of virtualized environments in cloud computing. The research work and results are of certain significance and scientific research value for the security of cloud computing virtualization environment.