L-Semi-Supervised Clustering for Network Intrusion Detection
Srinivasa Rao Narisetty1, Shaik Farzana2, Potnuri Maheswari3
1Srinivasa Rao Narisetty, Lakireddy Bali Reddy College of Engineering, Mylavaram, Krishna (Andhra Pradesh), India.
2Shaik Farzana, Lakireddy Bali Reddy College of Engineering, Mylavaram, Krishna (Andhra Pradesh), India.
3Potnuri Maheswari, Lakireddy Bali Reddy College of Engineering, Mylavaram, Krishna (Andhra Pradesh), India.
Manuscript received on 29 May 2019 | Revised Manuscript received on 11 June 2019 | Manuscript Published on 22 June 2019 | PP: 805-809 | Volume-8 Issue-3S, February 2019 | Retrieval Number: C11700283S19/19©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: To identify and detect network intrusion attack is a challenging problem in the network communication. The major problem with these attacks is that they can exploit the network vulnerabilities and steal the sensitive information from the organizations. These intruders use polymorphic approaches to masquerade their identity to detect. In recent times, many supervised and unsupervised Machine Learning algorithms have been proposed to detect network attacks. Supervised learning requires labeled information to build a classifier. Indeed it requires do-main experts to label each attack. These issues are addressed by semi-supervised learning (SSL) approach where it builds a classifier from few labeled datasets. This paper proposes a novel leader based SSL approach by using labeled and unlabeled patterns to improve the performance of Intrusion Detection Systems (IDS). It has two step approaches- the first step it derives a set of prototypes by using a fast-clustering method along with constraints called the constrained leaders clustering method with threshold parameter ζ. The second step is by applying the single link method in the presence of a few labeled data with respective constraints. The experimental results are obtained from the standard dataset NSL-KDD which is an extension of KDDCUP-99 datasets where the proposed constrained leader-based SSL method achieved better accuracy even with few labeled training patterns.
Keywords: Semi-Supervised, Intrusion Detection, Single Link Clustering, Machine Learning, Intrusion Prevention System.
Scope of the Article: Clustering