This paper presents a novel approach for training a network intrusion detection system based on a query-based sampling (QBS) filter. The proposed QBS filter applies the concepts of data quantization to signal processing in order to develop a novel classification system. Through interaction with a partially trained classifier, the QBS filter can use an oracle to produce high-quality training data. We tested the method with a benchmark intrusion dataset to verify its performance and effectiveness. Results show that selecting qualified training data will have an impact not only on the performance but also on overall execution (to reduce distortion). This method can significantly increase the accuracy of the detection rate for suspicious activity and can recognize rare attacks. Additionally, the method can improve the efficiency of real-time intrusion detection models.

To access the full article, please see PDF.

Authors and Affiliations

1. Department of Engineering Science and Ocean Engineering, National Taiwan University, Taipei, 10617, Taiwan

   Liang-Bin Lai, Ray-I Chang & Jen-Shiang Kouh

Corresponding author

Correspondence toRay-I Chang.

Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License (https://creativecommons.org/licenses/by/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions