Cyberthreat Detection Using Machine Learning

Millions of users have been a victim of cyberattacks, and thousands of companies are affected as well. This paper proposes Machine Learning to be used as a method to improve the detection rates of cyberthreats in a network which is better than the traditional signature or anomaly-based methods. Machine Learning can be used to detect threats and protect systems in real time thereby reducing the damage caused by attacks to a very high extent. In this paper, five Supervised Machine Learning algorithms, Random Forest, Logistic Regression, SVM, Decision Tree and Naive Bayes, have been used with optimized parameters and tuning and lastly, a deep learning algorithm; Convolutional Neural Network (CNN) has been used, and the performances have been compared among them. The algorithms performed well with Random Forest model being the highest. The results achieved prove that Machine Learning can be implemented to develop a threat detection system for a network which would be much more secure compared to the existing methods of detection and prevention.