An Efficient Machine Learning based Multiclass Cyber Attacks Classification and Prediction

Abstract

Security breach involves an attempt to attack a machine for data theft and compromising network resources to retrieve unauthorized information, network infrastructure, etc. The purpose of network attacks is to obtain information gathering, destroy network infrastructure and compromise data, etc. Cyberattack wields the deliberate contrivance of malicious code within the system’s logic and information to elicit unauthorized information through such hoodwinks resulting in cybercrime and fraud. This paper proposes a suitable classifier for cyberattack prediction based on an enhanced machine-learning technique. The study aims to identify and predict various techniques of cyberattacks such as Random Forest and XGBoost and other classifications for detection of attacks w.r.t. information available to the public.  According to the findings, Random Forest recall (RE) accuracy is 69% and precision (PR) is 69% for the initial classification that is Random Forrest. The F1 score of 69% is represented by the average accuracy. According to the report, the XGBoost model performed with precision (PR) of 75% and recall (RE) of 75% accuracy while using the second classification approach. Our model's average accuracy (AC) is 75%. Both models namely Random Forest and XGBoost show superiority over well-established methods for the classification and prediction of cyber-attacks.