Network Anomaly Uncovering on CICIDS-2017 Dataset: A Supervised Artificial Intelligence Approach
Pankaj Jairu, Akalanka B. Mailewa
Abstract
In today’s world, businesses and services are shifted to a digital transformation. As a result, network traffic has tremendously increased over the years. With that, network threats and attacks are growing and with that, the importance of intrusion detection systems has increased. The traditional signature-based approach to intrusion detection is not sufficient to detect intrusions, so anomaly-based intrusion detection came into play. There are many methods to Anomaly-based intrusion detection methods that can classify unknown network attacks. To detect network anomalies, Machine Learning and Deep Learning techniques are applied, and a considerable number of studies are done in this field. This research presents classification models built using supervised Machine Learning algorithms. The algorithms Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Naïve Bayes, Decision Tree and Random Forest on multiple datasets of realistic evaluation dataset CICIDS-2017. The results show that Random Forest outperforms other supervised algorithms with as high as 99.93% accuracy using 14 features selected using Pearson’s correlation coefficient method.