Title: Network intrusion detection using adversarial computational intelligence

Authors: Sudhir Kumar Pandey; Ditipriya Sinha

Addresses: Computer Science and Engineering, National Institute of Technology (NIT), Patna, 800005, Bihar, India ' Computer Science and Engineering, National Institute of Technology (NIT), Patna, 800005, Bihar, India

Abstract: Conventional intrusion detection systems (IDSs) in network ecosystems frequently face difficulties in recognising new types of attacks and navigating intricate network architectures, often resulting in a high false positives. Over the past recent years, researchers have probed a range of machine learning and deep learning frameworks to tackle these challenges, although many of these models demand more labelled data than what is typically accessible. To mitigate the data scarcity issue, researchers have begun utilising generative adversarial networks (GANs), specifically the external classifier generative adversarial network (EC-GAN) approach, to generate synthetic data. Our study employs a deep neural network (DNN) classifier, trained using EC-GAN, and benchmarks its performance against both earlier research and traditional training techniques. Remarkably, the classifier that leveraged the EC-GAN method demonstrated superior performance compared to other investigations, even though it required only a small portion of the original training dataset. To judge the computational novelty, comparative analysis and bench-marking is also carried out.

Keywords: GANs; generative adversarial networks; EC-GAN; external classifier generative adversarial network; DNNs; deep neural networks; semi-supervised learning; IDS; intrusion detection systems; CICIDS-2017.

DOI: 10.1504/IJCSM.2024.143207

International Journal of Computing Science and Mathematics, 2024 Vol.20 No.4, pp.290 - 307

Received: 28 Nov 2023
Accepted: 01 Jul 2024
Published online: 09 Dec 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article