An Intruder Detection System based on Feature Selection using Random Forest Algorithm
G. Madhukar1, G. Nantha Kumar2

1G. Madhukar, Research Scholar, Sri Satya Sai University of Technology & Medical Sciences, Sehore (M.P.), India.
2G. Nantha Kumar, Associate Professor, Sri Satya Sai University of Technology & Medical Sciences, Sehore (M.P.), India.
Manuscript received on November 20, 2019. | Revised Manuscript received on December 30, 2019. | Manuscript published on December 30, 2019. | PP: 5525-5529  | Volume-9 Issue-2, December, 2019. | Retrieval Number: B5154129219/2019©BEIESP | DOI: 10.35940/ijeat.B5154.129219
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (

Abstract: In every part of the world, there is tremendous growth in digital literacy in the present era. People are trying to access internet-based applications with the use of digital machines. As a result, the internet has become a primary requirement for everyone, and most business transactions often take place conveniently across the network. On the other hand, intruders involved in making intrusions and doing activities such as capturing passwords, compromise on the route, collecting details of credit cards, etc. Many malicious activities are taking place over the network due to this intruding activity on the internet. Applications such as host-based Intrusion Detection System (IDS) and network-based IDS have previously been used to control network intruders. Mostly when they come with Encrypted packets, spoofed network ids, these techniques were not able to control intruders promisingly. It is essential to examine these types of attacks periodically to identify patterns of recent attacks. In this paper, the authors have proposed a model based on deep learning by using the NSL – KDD dataset to solve these problems. For later train, the model with data with a random forest classifier algorithm, the principal component analysis applied for feature selection. The model is designed to detect patterns of intruders effectively using the knowledge gained from training data. To detect malicious patterns over the network, the model shows a sufficient accuracy of around 90 percent.
Keywords: Feature selection, Intrusion detection, Random forest, Principle component analysis, NSL-KDD dataset.