Modeling cloud computing risk assessment using ensemble methods

Abstract

Risk Assessment is a common practice in the information system security domain, besides that it is a useful tool to assess risk exposure and drive management decisions. Cloud computing has been an emerging computing model in the IT field. It provides computing resources as general utilities that can be leased and released by users in an on-demand fashion. It is about growing interest in many companies around the globe, but adopting cloud computing comes with greater risks, which need to be assessed. The main target of risk assessment is to define appropriate controls for reducing or eliminating those risks. The goal of this paper was to use an ensemble technique to increase the predictive performance. The main idea of using ensembles is that the combination of predictors can lead to an improvement of a risk assessment model in terms of better generalization and/or in terms of increased efficiency. We conducted a survey and formulated different associated risk factors to simulate the data from the experiments. We applied different feature selection algorithms such as best-first and random search algorithms and ranking methods to reduce the attributes to 4, 5, and 10 attributes, which enabled us to achieve better accuracy. Six function approximation algorithms, namely Isotonic Regression, Randomizable Filter Classifier, Kstar, Extra tree, IBK, and the multilayered perceptron, were selected after experimenting with more than thirty different algorithms. Further, the meta-schemes algorithm named voting is adopted to improve the generalization performance of best individual classifier and to build highly accurate risk assessment model.