Title
Embedding Time-Series Features into Generative Adversarial Networks for Intrusion Detection in Internet of Things Networks
Document Type
Article
Publication Date
1-1-2022
Publication Title
Intelligent Systems Reference Library
Volume
217
First Page
169
Last Page
183
Abstract
In recent years, Generative Adversarial Networks (GAN) have become powerful industrial tools to facilitate various learning tasks, including anomaly detection. This chapter studies a number of GAN architectures used for anomaly detection in the data stream. Moreover, a novel approach is proposed for embedding the dynamic characteristics of the data stream into the GAN-based detector structures. In this process, a GAN model is also proposed for efficient estimation of a confidence measure during the operation that reflects how well samples can be assigned to benign data. Furthermore, this chapter designs an intrusion detection system by developing a GAN-based anomaly detector. To do this, we study the effect of the proposed approach and the selected GAN-based approaches in detecting malicious intrusions in an Internet of Things (IoT) network. Experiments are evaluated in terms of false alarm and missed alarm detection rates. The obtained results indicate the effectiveness of the proposed GAN-based detection approach for the respective task.
DOI
10.1007/978-3-030-91390-8_8
ISSN
18684394
E-ISSN
18684408
Recommended Citation
Hallaji, Ehsan; Razavi-Far, Roozbeh; and Saif, Mehrdad. (2022). Embedding Time-Series Features into Generative Adversarial Networks for Intrusion Detection in Internet of Things Networks. Intelligent Systems Reference Library, 217, 169-183.
https://scholar.uwindsor.ca/electricalengpub/76