Artificial immunity based distributed and fast anomaly detection for Industrial Internet of Things
Beibei Li,Yujie Chang,Hanyuan Huang,Wenshan Li,Tao Li,Wen Chen
DOI: https://doi.org/10.1016/j.future.2023.06.011
IF: 7.307
2023-06-18
Future Generation Computer Systems
Abstract:Recent years have witnessed an increased attack surface of the Industrial Internet of Things (IIoT), as the deep convergence of the Internet of Things (IoT) and other information and communications technologies (ICTs). However, the massive geographically-dispersed IoT devices and the big spatiotemporal data they generated, both pose extreme challenges to traditional centralized and cumbersome anomaly detection systems (ADSs). To meet this gap, in this paper we propose a novel artificial immunity based distributed and fast anomaly detection system, coined AIm-ADS. Specifically, first we develop an artificial immune system based data representation and distributed preprocessing framework, to characterize and prepare the hyper-dimensional IIoT data for anomaly detection (i.e., distinguishing between selfs and nonselfs). Second, we design a new negative selection based hyper-dimensional anomaly detector generation model along with a point-hyperspace (PH) tree based self-detector indexing mechanism, for creating anomaly detectors and indexing self-detectors in a hyperspace. Last, we craft a PH tree based fast anomaly detection method to identify anomalous behaviors of cyberattacks on IIoT. Extensive experiments on three publicly available and widely used real-world datasets (i.e., SDS, SWaT and WADI), demonstrate that the proposed AIm-ADS outperforms existing artificial immunity based and deep learning based anomaly detection studies, in terms of both the detection rate and detection time.
computer science, theory & methods