Artificial Intelligence for Detection, Estimation, and Compensation of Malicious Attacks in Nonlinear Cyber Physical Systems and Industrial IoT

Faezeh Farivar, Mohammad Sayad Haghighi, Alireza Jolfaei, Mamoun Alazab

    Research output: Contribution to journalArticlepeer-review

    Abstract

    This paper proposes a hybrid intelligent-classic control approach for reconstruction and compensation of cyber attacks launched on inputs of nonlinear cyber-physical systems (CPS) and industrial IoT systems which work through shared communication networks. In this study, a class of n-order nonlinear systems is considered as a model of CPS while it is in presence of cyber attacks only in the forward channel. An intelligent-classic control system is developed to compensate cyber-attacks. Neural network (NN) is designed as an intelligent estimator for attack estimation and a classic nonlinear control system based on the variable structure control method is designed to compensate the effect of attacks and control the system performance in tracking applications. In the proposed strategy, nonlinear control theory is applied to guarantee the stability of the system when attacks happen. In this strategy, a Gaussian radial basis function NN is used for online estimation and reconstruction of cyber-attacks launched on the networked system. An adaptation law of the intelligent estimator is derived from a Lyapunov function. Simulation results demonstrate the validity and feasibility of the proposed strategy in car cruise control application as the testbed.
    Original languageEnglish
    Article number8917652
    Pages (from-to)2716-2725
    Number of pages10
    JournalIEEE Transactions on Industrial Informatics
    Volume16
    Issue number4
    Early online date28 Nov 2019
    DOIs
    Publication statusPublished - Apr 2020

    Fingerprint

    Dive into the research topics of 'Artificial Intelligence for Detection, Estimation, and Compensation of Malicious Attacks in Nonlinear Cyber Physical Systems and Industrial IoT'. Together they form a unique fingerprint.

    Cite this