CYBER ATTACKS DETECTION USING MACHINE LEARNING TECHNIQUES

Abstract

A reliable Cyber Attack Detection Model (CADM) is an invaluable tool for users of 
contemporary technological devices and network operators in their fight against 
cybercrime. Developing a CADM capable of accurately classifying cyber-attacks via 
pattern analysis of network data is the primary objective of this research. Using an 
ensemble classification approach, CADM determines the accuracy of attack-wise 
detection. When it comes to feature extraction, LASSO is the way to go for better 
visualisation and efficient processing of massive datasets. The ensemble method 
classifies network traffic data using algorithms like Gradient Boosting and Random 
Forest. Gradient Boosting optimises weak learning models to choose the best decision 
trees, while Random Forest trains all the trees at once. This increases prediction 
accuracy. The article also uses five datasets (NSL-KDD, KDD Cup 99, UNSWNB15, 
URL 2016, and CICIDS 2017) to evaluate the proposed model and verify that it is 
successful in ML-based cyberattack detection.