Comparative analysis of various ensemble classifiers using ensemble feature selection for detecting and preventing trespassers in hybrid cloud(PDECC)

We are in the era of anywhere any time computing and accessible to resources. Utility computing is the current and future of computing. However, at the same time cloud is vulnerable to attacks. We developed a Collaborative Intrusion Prevention and detection system in private multi cloud, in this one or more CSP’s are collaborated to deliver services to their registered users or clients. Our detection mechanism is done in a collaborated way if an anomaly is detected by one csp csp will inform other csp’s regarding this attack. Most of the earlier work is fo-cused on static data set i.e kdd cup 98, Darpa, cloud intrusion data set or on any other static data set. The earlier approaches of classification is processed on entire data set, or on feature-processed data or on correlation analysis of the attributes of the data set. In this paper we fo-cused on capturing live data using wire shark and for classification is done on based on closeness of the attributes called community which yields a better classification compared to correlation and other approaches. We recorded live traffic using Wire shark and pre-processing has been done using ensemble filtering techniques and detection has been done using Ensemble Classifiers Bagging, Boosting and Stacking and the results are compared. PDECC Bagging Random Forest Classifier achieved a high accuracy compared with PDECC Boosting, PDECC stacking and other methods.