Supervised Machine Learning for Cloud Security

Abstract

Although there is a lot of interest in cloud computing, security concerns have prevented it from becoming mainstream. Users of cloud services frequently worry about the loss of data, the compromise of security, and the unavailability of the services at important moments. Security applications that employ learning-based solutions are gaining attraction in the literature thanks to recent advancements. However, the most challenging aspect of these approaches is the objective datasets. Numerous internal datasets are off-limits for public usage for various reasons, including privacy and the possibility of missing statistical information. Even though there is some lacking, researchers are using these datasets for training and testing in experimental settings. Using a single dataset to train a machine learning model often produces misleading findings. How well these models perform when applied to data from a variety of sources and contexts is an open question, although it hasn’t been thoroughly explored in the literature. As, cloud problems are unique, therefore it is crucial to evaluate the performance of these models over a wide range of circumstances. To train the supervised machine learning models used in this research, we make use of the dataset made available by UNSW. For evaluating the performance of these models, we have used the ISOT dataset.