Statistical Model for Cybercrime Detection in Joined Heterogeneous Cloud Computing Environments

Abstract

Many commercial facilities and academic institutions now use cloud computing in an attempt of adopting new digital innovations. Cloud Computing Providers might be constrained to some services, lacking several resources their customers requested, meaning that distinct cloud services need to come together in collaboration to interoperate and exchange resources among themselves. In different characteristics and structures, clouds may be interlinked, and the interconnected systems may be prone to instability or intrusion. Although digital transformation and cloud services are also making progress in meeting the influencing policies internationally, terrorists use virtual space to commit cyberattacks. Adoption of cloud services has become highly sensitive to attacks and intrusions. Security breach or corruption of data gives organizations or agencies significant catastrophic losses. Divine influence and physical devices really aren't sufficient to protect services provided by clouds; thus, there is a need for an effective cyber protection model that is implementable, versatile, reliable and capable of detecting hazardous cyber-attacks on joined heterogeneous cloud providers, making it essential to decrease in real-time. This paper focuses on developing a Statistical Model for cybercrime detection in a heterogeneous cloud systems that are joined. When defining the implementation of the proposed model, we used an architecture and design modelling system. We used statistical notations and diagrams to contain the complexities and variability of the joining cloud data centres when adopting shared resources and possible cybercrime detection. The proposed model was experimented using WEKA tool to test Anomaly and Normal Accuracy performance of three (3) Classification existing algorithms: C4.5 and Random Decision tree and NaïveBayes. The dataset used was CICDDOS2019, the dataset was generated observing abstract behaviour of 25 users based on heterogeneous HTTP, HTTPS, FTP, SSH, and email protocols, Operating Systems and initiating different DDOS attacks. After a comparison of Classification model results, yielding in the highest accuracy and detection rates as well as the lowest false rates. The results specify that the classification capability of the C4.5 is inherently superior to the classification of Random Tree and the NaïveBayes.