Intrusion Detection System Using Ensemble Machine Learning in Cloud Environment

Abstract

Due to the expeditious evolution of advanced computing innovations, new organizational and operational strategies call for adoption. In the last few years, cloud computing has emerged and evolved as one such example to provide complex IT infrastructures, remote storage, a platform for developers and software according to user agreement. In order to make cloud computing possible which provides a distributed domain to its users, the internet has played a significant role. Irrespective of cloud’s versatility and efficiency, there are vulnerabilities present in the cloud environment which can arise both from within and outside the cloud. To protect clouds from any security risk, the traditional security methods fail to completely guard the networks and devices. As a consequence, building an effective intrusion detection system which can detect intrusions with higher accuracy in cloud environment needs to be prioritized. In this paper, we have suggested performing feature selection for reducing dimensions before applying ensemble method to build a model for detection of intrusions. A tree-based technique has been used to select the most prominent features. The intrusion detection system devised is anomaly-based that employs machine learning algorithms: Bagging algorithm and Random Forest algorithm on the UNSW-NB15 dataset which is more recent and comprehensive. An accuracy of 99.5% was achieved for both the bagging algorithm and Random Forest algorithm by reducing the dimension up to only 11 features.