Date of Award

Spring 2026

Abstract

The advanced manufacturing industry (AMI) faces unique challenges in security assurance. Security threats in AMI can originate from both software and hardware. Over the past decade, software security has been addressed extensively, but hardware security has not received enough attention. This work focuses on the security vulnerabilities of typical electronic devices deployed to AMI. Three attack models are proposed to characterize sensing nodes, local processing edge devices, and wired/wireless communication interfaces. Moreover, practical security attacks on hardware are demonstrated to inspire the development of feasible countermeasures against common physical attacks. To explore effective attack mitigation methods, we examine two attack scenarios, jamming and replay attacks, in the long-range wide-area networks (LoRaWAN) used in AMI. A unified defense framework is proposed to simultaneously detect and mitigate jamming and replay attacks conducted at multiple attack surfaces. To save the overhead on attack mitigation, an Advanced Continuous Time Convolution (ACTC) approach is implemented at the hardware level to obtain a high detection rate for jamming and replay attacks. We further compare our hardware-level attack mitigation with convolutional and recurrent neural network (CNN/RNN) based attack detection methods. Our experimental results indicate that the ACTC detection algorithm provides a balanced detection and classification accuracy in the 125 KHz and 500 KHz bandwidths.

Document Type

Dissertation

First Advisor

Qiaoyan Yu

Second Advisor

John LaCourse

Third Advisor

Dean Sullivan

Department or Program

Electrical and Computer Engineering

Degree Name

Doctor of Philosophy

Share

COinS