随着云计算技术的发展，商业云资源的使用成本越来越低，恶意用户可能利用云平台资源对同驻的虚拟机或者其他云平台实施入侵攻击。针对云服务的入侵攻击主要包括对虚拟机或监视器的攻击和后门通道攻击。针对现有云入侵检测系统仅能检测已知的攻击、对不同虚拟网络模型的兼容性较低、对攻击的变种的检测精度较低等问题，在分析KVM网络模型的基础上，提出一种云环境下基于改进BP算法的入侵检测模型( MBPCIDM)。该模型结合了PSO算法的全局寻优能力和BP算法的梯度下降局部搜索等特点，将PSO算法引入到BP的初始权值与阈值的优化，融入了动量项与自适应学习速率方法，使得BP网络更快收敛，且有效避免了算法陷入局部最优。实验结果表明，所提出的模型平均检出率较高，能为云环境提供入侵检测服务。%With the development of cloud computing technology,the cost of commercial cloud resources is lower and lower,a malicious user could use the cloud resources in the same virtual machine or other cloud platform to implement intrusion attack. The intrusion attack for cloud service mainly includes the virtual machine or monitor attack and back channel attack. The existing cloud intrusion detection systems can only detect known attacks,cannot be applied to a virtualized environment that has different network models,and the detection accuracy of variant of attack is lower. Based on the analysis of the KVM network structures,an improved intrusion detection model based on BP algorithm in the cloud environment ( MBPCIDM) was proposed. It combines the ability of searching global optimal solution of PSO algorithm and the feature of the gradient descent in local search of BP algorithm. To make the BP network convergence faster and prevent it from falling into local optimum,the momentum and adaptive learning rate method was also used in this paper. The experimental results show that the average detection rate of the proposed model is higher,it can provide intrusion detection services for cloud environ-ments.
展开▼
