Security Attack Detection Using Machine Learning and Deep Learning

Abstract

Data has been increasing exponentially over the past few years. This has introduced a variety of anomalies and security threats in the field of networking. Because of this, Intrusion Detection Systems (IDS) have become popular. An IDS is a network trafficking monitoring system that detects any suspicious activity and alerts the administrator when it is found. An IDS is generally prone to false alarms. Therefore, there is a need to fine-tune them. Using Machine Learning and Deep Learning techniques, one can build such systems which are accurate in detection of malicious activities. Previously, KDDCUP99 dataset has been used for most of the research purposes. However, the dataset is old and doesn’t justify the intrusions and malicious activities that occur now. Therefore, we have made the use of UNSW-NB15dataset (Moustafa and Slay, 2015 Military communications and information systems conference (MilCIS), 2015) which is an upgraded version of KDDCUP99 dataset and is more uniform and balanced. We have used various machine learning and deep learning algorithms like Logistic Regression (LR), Decision Trees (DT), Random Forest (RF), Naïve Bayes (NB), Support Vector Machine (SVM) and Artificial Neural Networks (ANN) to classify whether it was a network intrusion or not. It has been found that SVM, DT, ANN and NB gave a hundred percent accuracy whereas the Random Forest model gave an accuracy of 99.9%.