The Analysis Methods Of Data Fusion For Cloud Monitoring

With cloud security issues becoming increasingly prominent,the demand for reasonable cloud security monitoring,assessment and prediction system is increasingly urgent.While,the analysis technology of data fusion for cloud which is an important part of this system can provide evidence directly to support cloud security monitoring system.However,the cloud environment has a complex structure and much more heterogeneity data with time,space and semantic.The traditional analysis technology of data fusion is not well adapted to these characteristics,which seriously hindered the development of cloud security surveillance.Traditional data fusion technology is divided into three levels,the principles and algorithms of each level are not same,which can be used either in combination or single.However,the former is excessive and overhead,the latter has low accuracy.They are inappropriate cloud monitoring system.Accordingly,this paper proposes an improved two-level data fusion model with data-level and decision-level,which makes up for the previous lack and makes the results more accurate.The traditional association rule algorithms use serial mode to analysis,the low efficiency of implementation,the generation of candidate sets,and the memory usage,making them not suitable for cloud monitoring system.Paper presents an improved parallel Apriori algorithm based on Map-Reduce.Firstly,it is pretreated by making the original database into Boolean project database,and stored by bitmap method;Secondly,Hash table is used to divide the data and makes the same IP distributed into same block;Thirdly,the Map function is improved by using multi-tree structure to store frequent item sets,and merge the multi tree with Reduce function;Finally,this algorithm is verified with higher efficiency and accuracy.The traditional methods have low efficiency to generate game model,this paper presents a parallel algorithm for stochastic game model.Firstly,according to the topology,we generate stochastic petri net attack model,and have a stochastic attack-defense game model combined with defense strategy.Secondly,we calculate the revenue function,select the high–yielding and stable income strategy by analysising the contribution rate in information entropy.Finally,we can accuratelygive the probability of strategy selection through experiments,and get the performance of the cloud system.In summary,this paper proposes a two-level data fusion analysis model for big data,Which includes the parallel Apriori algorithm and stochastic game model to achieve security analysis.Simulation results show that,compared with the traditional method,this method has lower overhead and higher accuracy.