Adapting Noise-Driven PUF and AI for Secure WBG ICS: A Proof-of-Concept Study

Devon A. Kelly, Christiana Chamon · 2025

Read Adapting Noise-Driven PUF and AI for Secure WBG ICS: A Proof-of-Concept Study by Devon A. Kelly, Christiana Chamon (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-22283.txt.

Introduction

Devon A. Kelly and Christiana Chamon turn an industrial nuisance into a security asset. Wide-bandgap power electronics switch fast and efficiently, but their high-frequency noise corrupts sensors and opens cyber-physical attack surface inside industrial control systems. The team's proof of concept harvests that very noise, up to one hundred kilohertz, as a source of physical identity.

A physically unclonable function extracts entropy from the unavoidable switching noise, giving each device a hardware fingerprint that cannot be copied. The same noise stream doubles as a live threat indicator. Hybrid machine learning models with adaptive Bayesian filtering watch for anomalies at low latency, tuned to stay robust under natural electromagnetic interference as well as attack. Authentication and anomaly detection thus share one signal, a neat fit for sensor pathways that are too noisy for conventional security.

Catalog record

CREATORDevon A. Kelly, Christiana Chamon
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-22283

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Adapting Noise-Driven PUF and AI for Secure WBG ICS: A Proof-of-Concept Study Devon A. Kelly, Christiana Chamon Abstract Wide-bandgap (WBG) technologies offer unprecedented improvements in power system efficiency, size, and performance, but also introduce unique sensor corruption and cybersecurity risks in industrial control systems (ICS), particularly due to high-frequency noise and sophisticated cyber-physical threats. This proof-of-concept (PoC) study demonstrates the adaptation of a noise-driven physically unclonable function (PUF) and machine learning (ML)-assisted anomaly detection framework to the demanding environment of WBG-based ICS sensor pathways. By extracting entropy from unavoidable WBG switching noise (up to 100 kHz) as a PUF source, and simultaneously using this noise as a real-time threat indicator, the proposed system unites hardware-level authentication and anomaly detection. Our approach integrates hybrid machine learning (ML) models with adaptive Bayesian filtering, providing robust and low-latency detection capabilities resilient to both natural electromagnetic interference (EMI) and active adversarial manipulation. Through detailed simulations of WBG modules under benign and attack scenarios—including EMI injection, signal tampering, and node impersonation—we achieve 95% detection accuracy and sub-millisecond processing latency. These results demonstrate the feasibility of physics-driven, dual-use noise exploitation as a scalable ICS defense primitive. Our findings lay the groundwork for next-generation security strategies that leverage inherent device characteristics, bridging hardware and artificial intelligence (AI) for enhanced protection of critical ICS infrastructure. I Introduction I-A Problem Statement Industrial control systems (ICS) form the technological backbone of modern critical infrastructure, governing processes in domains such as electric power generation and transmission, manufacturing, oil and gas, transportation, and water treatment. These systems encompass Supervisory Control and Data Acquisition (SCADA) networks, Programmable Logic Controllers (PLCs), and a variety of sensory and actuation devices orchestrated under time-critical, high availability constraints. Traditionally, the security posture of ICS has focused on reliability and fault tolerance; however, the increasing interconnection with enterprise networks, and more recently with the Industrial Internet of Things (IIoT), has exposed them to a growing class of cyber-physical threats [ 2 , 3 , 1 ] . Among the technological advances shaping the next generation of ICS is the integration of wide-bandgap (WBG) semiconductor power devices, such as silicon carbide (SiC) and gallium nitride (GaN). These devices provide superior efficiency, faster switching transients, and higher power density compared to their silicon counterparts [ 5 , 4 ] . As utilities and industrial operators pursue aggressive decarbonization and efficiency initiatives, WBG-based modules are being rapidly adopted in motor drives, renewable energy converters, and high-voltage direct current (HVDC) grids. While these capabilities significantly enhance operational performance, they also introduce unique security challenges. The fast switching actions intrinsic to WBG devices generate electrical noise and electromagnetic interference (EMI) at frequencies often exceeding 100 kHz. This spectral behavior can compromise sensor fidelity, interfere with communication channels, and—if maliciously manipulated—undermine system stability. Emerging threats in this space have demonstrated the vulnerability of ICS to both accidental and adversarial corruption of sensor data. False data injection attacks, for instance, can compromise state estimation and lead operators to incorrect control decisions [ 6 , 7 ] . When combined with the inherent noise characteristics of WBG power stages, the attack surface broadens. Malicious actors could exploit EMI or high-frequency perturbations to spoof measurements, degrade anomaly detection systems, or mask physical tampering. This duality—where noise is both a natural byproduct of efficient power electronics and a potential weapon for adversaries—necessitates novel defensive strategies that are specifically tailored to the realities of WBG-enhanced ICS environments. One promising defense primitive lies in the concept of physically unclonable functions (PUFs) [ 8 , 9 , 10 , 11 ] . In hardware security research, PUFs have been successfully applied to device authentication, cryptographic key generation, and tamper detection. However, the majority of these implementations assume traditional entropy sources such as delay variations in ring oscillators or startup states of SRAM cells. Little work has explored leveraging the unavoidable switching noise in power systems as an entropy source for PUFs—particularly in the context of ICS. In parallel, machine learning (ML) methods have risen in prominence as enablers of adaptive anomaly detection in ICS and IIoT networks [ 12 , 13 ] . Autoencoders, reinforcement learning (RL), and Bayesian inference techniques have been deployed to monitor telemetry data, learn normal operating profiles, and flag deviations in real-time. Despite their success, these solutions are often vulnerable to adversarial input perturbations and require adaptation to the unique latency and reliability constraints of control systems. This work proposes a novel integration of these two research directions: coupling a noise-driven PUF derived from WBG switching spectra with an artificial intelligence (AI)-based anomaly detection framework . By treating high-frequency switching noise as both a challenge and an opportunity, the design provides two complementary security benefits. First, the noise characteristics serve as a physically grounded entropy source, producing unique identifiers for sensor and module authentication. Second, fluctuations in noise profiles can be continuously monitored using adaptive ML models, where deviations from expected PUF patterns indicate potential tampering or environmental anomaly. In effect, the same physical phenomenon that complicates system operation is recycled into a measurable, security-enhancing signal. The key contributions of this proof-of-concept study are as follows: • We introduce the concept of a noise-driven PUF for ICS, leveraging inherent switching noise in WBG modules as a unique and unclonable signature to authenticate sensor nodes. • We develop a hybrid anomaly detection framework integrating machine learning with Bayesian filtering, enabling responsiveness under sub-millisecond time constraints required by ICS loops. • We evaluate the feasibility of this approach in a controlled simulation environment that models WBG-based power electronics modules subject to EMI, malicious injection, and data corruption. • We present initial results suggesting that the approach can achieve 95% detection accuracy with processing latency below 1 ms, demonstrating its potential suitability for real-time ICS defense. I-B Motivation and Societal Impact Critical infrastructure sectors including electric utilities, water treatment, and advanced manufacturing are increasingly dependent on reliable and cyber-resilient ICS [ 1 ] . Recent high-profile ICS incidents such as the Colonial Pipeline ransomware attack and the Ukraine power grid compromise underscore the critical need for resilient sensor authentication mechanisms [ 14 ] . Attacks that manipulate sensor data threaten stability, safety, and reliability, motivating approaches that embed hardware-rooted, physics-driven security primitives. WBG-based power conversion technologies, promoted for grid flexibility and decarbonization, are rapidly deployed in microgrids, renewables, and HVDC interconnects. Their efficiency advantages come at the cost of new, little-studied attack surfaces due to high-frequency emissions and susceptibility to cyber-physical manipulation. The remainder of the paper is organized as follows. Section II surveys related work in ICS security, WBG-induced noise challenges, and PUF applications. Section III formalizes our system and threat models. Section IV details the design of the noise-driven PUF and the hybrid ML framework. Section V reports the proof-of-concept evaluation and performance results. Section VI provides insight on the feasibility of noise-driven PUFs for ICS. Section VII provides a discussion of our findings, limitations, and potential extensions. Finally, Section VIII concludes the paper with directions for future work. II Background and Related Work This section provides the necessary context to position our contribution. We first review security challenges in ICS and high-profile attack campaigns. We then introduce the role of WBG power electronics and their EMI

Read the complete text