云环境下拟态应用行为预测方法研究  

Research on prediction method of mimic application behavior in cloud environment

作　　者：徐悦 倪明[1] 余新胜[1] 解维[1] XU Yue;NI Ming;YU Xin-sheng;XIE Wei(The 32nd Research Institute of China Electronics Technology Group Corporation,Shanghai 201806,China)

机构地区：[1]中国电子科技集团公司第三十二研究所,上海201806

出　　处：《信息技术》2021年第1期53-59,共7页Information Technology

基　　金：上海市科学技术委员会建设基金(17DZ2251400)。

摘　　要：面对一些成熟的商业云,装配拟态服务组件可以为云上应用赋能内生安全。为了能够提供多应用与拟态服务组件集群的协同,文中首先提出并设计了拟态调度管理组件,通过高效调度拟态服务组件集群来满足拟态服务组件与云平台的融合协同需求。为了进一步优化所提出的拟态调度管理组件,文中同时研究了拟态调度管理组件中的预测算法,使用擅于处理线性关系的ARI-MA模型与擅于处理非线性关系的RBF模型组合,通过离线方式实现对云上应用行为预测。经过大量仿真实验,证明采用的组合模型的预测性能优于单个预测模型使用,能够以较高的准确率进行预测。For some mature commercial clouds,mimic service components can be assembled to provide endogenous security empowerment for cloud applications.In order to provide the collaboration of multi-application and mimic service component clusters,this paper first propose and designs the mimic scheduling management components,which can meet the requirements of cloud platforms through the efficient scheduling of mimic service component clusters.For further optimizing the proposed mimic scheduling management components,we also studies the prediction algorithm in the mimic scheduling management components.ARIMA model is good at handling linear relationships and RBF model is superior in handling nonlinear relationships.Using the combination of the ARIMA and RBF model,the behavior prediction of cloud applications can be realized in an offline manner.A large number of simulation experiments prove that the prediction performance of the combined model is better than the single model.

关 键 词：拟态防御 动态异构冗余 云平台 ARIMA模型 RBF模型 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]