Cyber attacks detection and type prediction for cloud system using machine learning techniques

Cyber security and Cloud platforms are utilized in various usage and applications in today’s world. Given the wide range of applications, and the ease of usage they provide, the popularity of them are increasing dramatically. Leading many individuals and organizations to depend on them mainly. Securing data, hardware, networks and other resources from cyber-attacks represent a crucial factor for these organizations. The work in this paper proposes an approach of multiple stages to detect and predict the cyber -attacks types aiming to enforce higher security procedures to secure the organization resources in general and data in specific. The approach first stage is the data collection where Meraz dataset available on the internet is used, and then different levels of preprocessing were conducted. The third stage is to apply different classification algorithms to group the attacks into malicious or not. Then after, the data related to the classifier that yield optimum classification results is selected for next level of knowledge extraction where hierarchical clustering was applied. The clustering is built on the malware samples of test dataset only. This dataset is divided into training and testing samples. A 10% of the dataset was used to predict the malware type. Hierarchical clustering was used with various configurations. The reason for using clustering is to predict the attack type by assigning each attack for distinct cluster. The proposed approach gave 98.88% of accuracy with Random Forest classifier and a reliable result for clustering were using Hierarchical clustering by using Euclidean distance metric, and ward linkage, The prediction values were as follows {0: 10671, 1: 3603, 2: 824}. The results obtained gave a novel approach for developing Machine Learning solution for cloud systems security. With this novel solution, the limitations of the traditional solutions are solved.