Design and Development of an ML/DL Attack Resistance of RC-Based PUF for IoT Security
Summary: arXiv:2603.28798v1 Announce Type: cross
Abstract
Physically Unclonable Functions (PUFs) provide promising hardware security for IoT authentication, leveraging inherent randomness suitable for resource-constrained environments. However, ML/DL modeling attacks threaten PUF security by learning challenge-response patterns. This work introduces a custom resistor-capacitor (RC) based dynamically reconfigurable PUF using 32-bit challenge-response pairs (CRPs) designed to resist such attacks.
Introduction
The growth of the Internet of Things (IoT) has led to an increased demand for effective security measures to protect devices and data. PUFs, which exploit the inherent manufacturing variations in hardware components, have emerged as a viable solution for secure device authentication. However, the rise of machine learning (ML) and deep learning (DL) techniques poses significant threats to the integrity of PUFs. This article discusses a novel RC-based PUF designed to enhance security against these advanced modeling attacks.
Methodology
The proposed RC-PUF utilizes a dynamically reconfigurable architecture to generate 32-bit challenge-response pairs (CRPs). The system’s robustness was evaluated by creating a comprehensive dataset of CRPs, which was subsequently divided into training, validation, and test sets. Various ML techniques were employed to assess the PUF’s vulnerability to modeling attacks, including:
- Artificial Neural Networks (ANN)
- Gradient Boosted Neural Networks (GBNN)
- Decision Trees (DT)
- Random Forests (RF)
- XGBoost
Results
All models achieved 100% training accuracy, indicating that they could perfectly learn the training data. However, the test performance revealed significant vulnerabilities, with results approximating random guessing:
- ANN: 51.05%
- GBNN: 53.27%
- DT: 50.06%
- RF: 52.08%
- XGBoost: 50.97%
These findings underscore the effectiveness of the proposed RC-PUF in resisting ML-driven modeling attacks, as even sophisticated algorithms struggled to replicate accurate responses.
Conclusion
The dynamically reconfigurable nature of the RC-PUF enhances its robustness against adversarial threats while maintaining minimal resource overhead. This innovative approach offers a low-cost alternative to more complex encryption methods, providing essential security for next-generation IoT authentication systems. By ensuring reliable device verification without compromising computational efficiency or scalability, this RC-PUF represents a significant advancement in the ongoing battle against machine learning-based threats.