UNSW-NB15 Dataset Feature Selection and Network Intrusion Detection using Deep Learning
V. Kanimozhi1, Prem Jacob2
1V. Kanimozhi, School of Computing, Sathyabama Institute of Science & Technology, Chennai (Tamil Nadu), India.
2Dr. Prem Jacob, School of Computing, Sathyabama Institute of Science & Technology, Chennai (Tamil Nadu), India.
Manuscript received on 08 February 2019 | Revised Manuscript received on 21 February 2019 | Manuscript Published on 04 March 2019 | PP: 443-446 | Volume-7 Issue-5S2 January 2019 | Retrieval Number: ES208001751919/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Anomaly detection system in network, monitors and detects intrusions in the networking area, which is referred to as NIDS, the Intrusion Detection System in Networks. There are numerous network datasets available in networking communications with relevant and irrelevant features drastically decreases the rate of intrusion detection and increases False Alarm Rate. The benchmark network dataset available is UNSW-NB15 dataset was created in 2015. The top significant features are proposed as feature selection for dimensionality reduction in order to obtain more accuracy in attack detection and to decrease False Alarm Rate. We apply a combination fusion of Random Forest Algorithm with Decision Tree Classifier using Anaconda3 (free and open-source distribution of Python3) and package management system Conda in which 45 features have been decreased to the strongest four features. The proposed system detects normal and attacks with a better accuracy using Deep Learning technique.
Keywords: Data Visualization; Feature Selection; Intrusion Detection; Artificial Neural Network; UNSW-NB15 Dataset.
Scope of the Article: Deep Learning