Session: 03-16-01: Securing Advanced Manufacturing: Cybersecurity and Edge Computing for Industrial IoT

Paper Number: 166056

Dynamic Watermarking for Cybersecurity of Machine Tool Controllers 

Modern manufacturing systems have become increasingly complex and interconnected, making them one of the most challenging domains for cybersecurity. Among the various components of a manufacturing system, Machine Tool Controllers (MTCs) represent a critical cybersecurity risk due to their role in governing the operation of machinery on the plant floor. A successful cyberattack on an MTC could lead to severe consequences, including operational disruptions, reduced productivity, compromised product integrity, and even safety hazards for human operators. Despite the growing concerns surrounding the cybersecurity of manufacturing networks, there remains a significant gap in real-time detection and mitigation techniques for securing MTCs against sophisticated cyber threats.

In recent years, Dynamic Watermarking (DWM) has emerged as a promising approach for detecting cyber intrusions in safety-critical systems. Originally developed for securing energy networks and autonomous vehicles, DWM provides strong theoretical guarantees by actively injecting known signals into the system and continuously monitoring the response to detect inconsistencies indicative of an attack. Given its success in other domains, this work presents the first adaptation of DWM for securing manufacturing MTCs. The proposed approach is designed to identify cyberattacks in real-time, enabling swift countermeasures to protect MTCs from potential sabotage while maintaining the system’s operational integrity.

One of the primary challenges in applying DWM to MTCs is the inherent non-linearity of these systems. Many real-world manufacturing processes involve complex dynamics that are not easily captured by standard linear models. To address this, we employ a piecewise linear system identification approach, wherein the MTC’s behavior is approximated through a set of local linear pieces, structured as AutoRegressive with eXogenous input (ARX) models of order (1,1). By dynamically switching between these models based on operating conditions, we achieve a more accurate representation of the system’s response while preserving the ability of DWM to detect cyber anomalies effectively.

To evaluate the effectiveness of DWM in manufacturing environments, this study considers both open and closed MTC architectures. The open MTC is represented by a smart stepper motor setup, which is physically implemented in a linear motion system. In contrast, the closed MTC is exemplified by the Siemens Sinumerik 828D controller, a widely used industrial-grade CNC controller. Due to the restrictive nature of closed MTC architectures, a Digital Twin (DT) of the Siemens controller is employed for extensive experimentation. DT allows for the simulation of cyberattacks and the assessment of detection performance in a controlled yet realistic environment.

The experimental results demonstrate the efficacy of DWM in detecting cyberattacks while accounting for system non-linearity. By leveraging piecewise linear approximations, the method successfully adapts to the complex dynamics of MTCs, ensuring reliable attack detection across different operational conditions. These results highlight the potential of DWM as a robust cybersecurity measure for MTCs, offering real-time attack detection with minimal impact on system performance. Future research will focus on refining the implementation for industrial deployment and extending its applicability to a broader range of manufacturing systems.

Presenting Author: Abhishek Hanchate Texas A&M University

Presenting Author Biography: Abhishek (Abhi) Hanchate, a 2024 SME 30 Under 30 Honoree, holds a bachelor’s degree from India’s COEP Technological University and a master’s from Texas A&M University. During his undergraduate studies, he conducted research on meta-heuristic optimization techniques as an exchange student at Nanyang Technological University (NTU), Singapore. Currently pursuing a Ph.D. at Texas A&M, he focuses on advancing smart manufacturing through machine learning, signal processing, and cybersecurity innovations.

Abhi's professional journey includes roles as a Data Scientist and Machine Learning Intern at American Airlines (Fall 2024), where he developed a personalized recommendation system for the AAdvantage Loyalty Program, and as Head AI Mentor at MIT FutureMakers (Summer 2024), guiding students in cutting-edge deep learning projects. Earlier, he leveraged data science to support the Texas A&M Aggies Football team in 2020.

Abhi has authored 8 peer-reviewed publications and holds 2 pending patents. In 2023, he founded AlignAI, a startup specializing in multimodal data fusion, securing $5K in seed funding. He actively contributes to the academic community as a reviewer for journals like JIMS and conferences such as NeurIPS and SME NAMRC/MSEC. A member of INFORMS, IISE, and SME, his manufacturing research excellence earned him the prestigious 2025 Dr. Milden J. Fox Jr. ’69 and Mary P. Fox ’73 Fellowship.

Authors:

Abhishek Hanchate Texas A&M University
Akash Tiwari Oak Ridge National Laboratory
Satish Bukkapatnam Texas A&M University