UTTAMA: An Intrusion Detection System Based on Feature Clustering and Feature Transformation

Detecting Intrusions and anomalies is becoming much more challenging with new attacks popping out over a period of time. Achieving better accuracies by applying benchmark classifier algorithms used for identifying intrusions and anomalies have several hidden data mining challenges. Although neglecte...

Full description

Saved in:
Bibliographic Details
Published in:Foundations of science 2020-12, Vol.25 (4), p.1049-1075
Main Authors: Nagaraja, Arun, Uma, B., Gunupudi, Rajesh kumar
Format: Article
Language:English
Subjects:
Algorithms
Anomalies
Classifiers
Clustering
Computation
Data mining
Datasets
Detectors
Education
Intrusion detection systems
Mathematical Logic and Foundations
Methodology of the Social Sciences
Philosophy
Philosophy of Science
Security software
Similarity
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Detecting Intrusions and anomalies is becoming much more challenging with new attacks popping out over a period of time. Achieving better accuracies by applying benchmark classifier algorithms used for identifying intrusions and anomalies have several hidden data mining challenges. Although neglected by many research findings, one of the most important and biggest challenges is the similarity or membership computation. Another challenge that cannot be simply neglected is the number of features that attributes to dimensionality. This research aims to come up with a new membership function to carry similarity computation that can be helpful for addressing feature dimensionality issues. In principle, this work is aimed at introducing a novel membership function that can help to achieve better classification accuracies and eventually lead to better intrusion and anomaly detection. Experiments are performed on KDD dataset with 41 attributes and also KDD dataset with 19 attributes. Recent approaches CANN and CLAPP have showed new approaches for intrusion detection. The proposed classifier is named as UTTAMA. UTTAMA performed better to both CANN and CLAPP approaches w.r.t overall classifier accuracy. Another promising outcome achieved using UTTAMA is the U2R and R2L attack accuracies. The importance of proposed approach is that the accuracy achieved using proposed approach outperforms CLAPP, CANN, SVM, KNN and other existing classifiers.
ISSN:1233-1821
1572-8471
DOI:10.1007/s10699-019-09589-5