Intrusion Detection System using Hybrid SVM-RF and SVM-DT in Wireless Sensor Networks
S. Prithi1, S. Sumathi2
1S. Prithi, Assistant Professor, Department of CSE, Rajalakshmi Engineering College, Chennai (Tamil Nadu), India.
2Dr. S. Sumathi, Professor, Department of EEE, PSG College of Technology, Coimbatore (Tamil Nadu), India.
Manuscript received on 25 August 2019 | Revised Manuscript received on 11 September 2019 | Manuscript Published on 17 September 2019 | PP: 1926-1931 | Volume-8 Issue-2S8 August 2019 | Retrieval Number: B12000882S819/2019©BEIESP | DOI: 10.35940/ijrte.B1200.0882S819
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: Intrusion detection system (IDS) is one of the essential security mechanisms against attacks in WSN. Network intrusion detection system (NIDS) generally uses the classification techniques in order to obtain the best possible accuracy and attack detection rate. In this paper, Intrusion Detection System is designed which uses two-stage hybrid classification method. In the first stage it uses Support Vector Machine (SVM) as anomaly detection, and in the second stage it uses Random Forest (RF)/Decision Tree (DT) as misuse. The abnormal activities are detected in the first stage. These abnormal activities are further analyzed and the known attacks are identified in the second stage and are classified as Denial of Service (DoS) attack, Probe attack, Remote to Local (R2L) attack and User to Root (U2R) attack. Simulation results reveal that the proposed hybrid algorithm obtains better accuracy and detection rate than the single classifier namely, SVM, RF and DT algorithm. The experimental results also shows that hybrid algorithm can detect anomaly activity in a reliable way. Proposed technique uses the standard NSL KDD dataset to evaluate/calculate the performance of the proposed approach. Here the results show that the proposed Hybrid SVM-RF/DT IDS technique performs better in terms of detection rate, accuracy and recall than the existing SVM, RF and DT approaches.
Keywords: Intrusion Detection System, Network Attacks, Support Vector Machine, Wireless Sensor Network.
Scope of the Article: Wireless ad hoc & Sensor Networks