Anomaly Analysis for the Classification Purpose of Intrusion Detection System with K-Nearest Neighbors and Deep Neural Network

Nowadays, along with network development, due to the threats of unknown sources, information communication is more vulnerable and require more secured information. An Intrusion Detection System (IDS) is important for protecting information with growing of unauthorized activities in-network. Traditional firewall techniques are less capable to protect information against new intrusion. Numerous researches on intrusion detection system have been conducted but old dataset like Kddcup'99 is analyzed. Problem identified that lack of accuracy to detect intrusion with the current available intrusion system. Hence this study aims to anomaly analysis for the classification purpose of the intrusion detection system with the most update dataset named CICIDS-2017 which can be used for the intrusion detection evaluation. This research has conducted the anomaly analysis for the classification purpose based on the K-Nearest Neighbors (KNN) for the machine learning (ML) and Deep Neural Network (DNN) using the Deep Learning (DL) method. One of the results presents a classification performance based on Matthews Correlation Coefficient (MCC) for ML and DL. DNN has performed significantly higher correctness classifier which shows DNN score 0.9293% compared to KNN is at 0.8824%. This research is significant as reference for IDS development which would improve security response for networked systems.