cs.CRMar 26, 2026

Design and Development of an ML/DL Attack Resistance of RC-Based PUF for IoT Security

Authors: Joy Acharya, Smit Patel, Paawan Sharma, Mohendra Roy

Organizations: Department of ICT, Pandit Deendayal Energy University Gandhinagar, Gujarat 382007, India

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

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. We systematically evaluated robustness by generating a CRP dataset and splitting it into training, validation, and test sets. Multiple ML techniques including Artificial Neural Networks (ANN), Gradient Boosted Neural Networks (GBNN), Decision Trees (DT), Random Forests (RF), and XGBoost, were trained to model PUF behavior. While all models achieved 100% training accuracy, test performance remained near random guessing: 51.05% (ANN), 53.27% (GBNN), 50.06% (DT), 52.08% (RF), and 50.97% (XGBoost). These results demonstrate the proposed PUF's strong resistance to ML-driven modeling attacks, as advanced algorithms fail to reproduce accurate responses. The dynamically reconfigurable architecture enhances robustness against adversarial threats with minimal resource overhead. This simple RC-PUF offers an effective, low-cost alternative to complex encryption for securing next-generation IoT authentication against machine learning-based threats, ensuring reliable device verification without compromising computational efficiency or scalability in deployed IoT networks.

Figures & tables

Fig. 1: RC-PUF dataset generation workflow showing 32-bit challenge input, analog delay measurement, and digital response storage.
Error FunctionEquation
Binary Cross-Entropy (BCE)LBCE=−132N∑i=1N∑j=132[yijlog⁡p^ij+(1−yij)log⁡(1−p^ij)]\mathcal{L}_{\text{BCE}}=-\frac{1}{32N}\sum_{i=1}^{N}\sum_{j=1}^{32}\big[y_{ij}\log\hat{p}_{ij}+(1-y_{ij})\log(1-\hat{p}_{ij})\big]
BCEWithLogitsLossLBCE+Logits=−132N∑i=1N∑j=132[yijlog⁡σ(z^ij)+(1−yij)log⁡(1−σ(z^ij))]\mathcal{L}_{\text{BCE+Logits}}=-\frac{1}{32N}\sum_{i=1}^{N}\sum_{j=1}^{32}\big[y_{ij}\log\sigma(\hat{z}_{ij})+(1-y_{ij})\log(1-\sigma(\hat{z}_{ij}))\big]
TABLE I: Error functions used in training.
MetricEquation
Bitwise AccuracyAccbit=132N∑i=1N∑j=132I[y^ij=yij]\text{Acc}_{\text{bit}}=\tfrac{1}{32N}\sum_{i=1}^{N}\sum_{j=1}^{32}\mathbb{I}[\hat{y}_{ij}=y_{ij}]
Hamming LossHL=132N∑i=1N∑j=132I[y^ij≠yij]\text{HL}=\tfrac{1}{32N}\sum_{i=1}^{N}\sum_{j=1}^{32}\mathbb{I}[\hat{y}_{ij}\neq y_{ij}]
Exact Match AccuracyEM=1N∑i=1NI[y^i=yi]\text{EM}=\tfrac{1}{N}\sum_{i=1}^{N}\mathbb{I}[\hat{\mathbf{y}}_{i}=\mathbf{y}_{i}]
TABLE II: Evaluation metrics for PUF modeling.
ModelTrain Accuracy (%)Val Accuracy (%)Exact Match (%)
ANN10051.060.00
GBNN10053.270.00
XGBoost10050.970.00
DT10050.370.00
RF10051.320.00
TABLE III: Model performance summary.
EntityMeanVarAvg EntropyMin Ent.Max Ent.
Target Y0.55010.24740.99260.98780.9983
ANN0.55840.24650.98980.98380.9960
GBNN0.54430.15160.67010.05440.9999
XGBoost0.55710.24670.99030.98340.9945
DT0.54660.24770.99350.98600.9969
RF0.58190.24310.98000.96000.9900
TABLE IV: Bit-level statistical analysis of predicted responses.
Fig. 2: Normalised step vs. training accuracy for all ML/DL models, showing convergence toward near-perfect training accuracy across ANN, GBNN, XGBoost, Decision Tree, and Random Forest.
Fig. 3: Normalised step vs. validation accuracy for all ML/DL models, showing fluctuations close to the random-guess baseline across ANN, GBNN, XGBoost, Decision Tree, and Random Forest.
Fig. 4: Comparison of test accuracy across different ML/DL models for RC-PUF modeling attacks.
PUFModelTrain (%)Test (%)Resistance
Arbiter PUF [ 12 ]LR/SVM∼\sim 100>> 90Low
XOR Arbiter PUF [ 12 ]ANN∼\sim 10070–85Moderate
RO PUF [ 13 ]RF/ANN∼\sim 10060–75Moderate
FF Arbiter PUF [ 14 ]SVM∼\sim 10065–80Moderate
Proposed RC-PUFMulti-ML/DL10050–53High
TABLE V: Comparison of ML/DL Attack Resistance with Existing PUFs
More in this topic

Internet of Things

The Internet of Things (IoT) connects networked sensors, devices, and systems for sensing, communication, and automated action.

See all 105 papers

Explore similar work

CardsList

  1. Robustness Analysis of Machine Learning Models for IoT Intrusion Detection Under Data Poisoning Attacks

    Apr 15, 2026Fortunatus Aabangbio Wulnye, Justice Owusu Agyemang, Kwame Opuni-Boachie Obour Agyekum +3IoT Intrusion DetectionAdversarial Robustness

    Department of Telecommunication, Kwame Nkrumah University of Science and Technology (KNUST), Kumasi, Ghana

  2. Toward a Multi-Layer ML-Based Security Framework for Industrial IoT

    Mar 25, 2026Aymen Bouferroum, Valeria Loscri, Abderrahim BenslimaneInternet of ThingsIoT Intrusion Detection

    LIA · Inria Lille-Nord Europe, Lille, France · LIA/CERI University of Avignon, Avignon, France

  3. A PUF-Based Approach for Copy Protection of Intellectual Property in Neural Network Models

    Mar 11, 2026Daniel Dorfmeister, Flavio Ferrarotti, Bernhard Fischer +2Cybersecurity

    Software Competence Center Hagenberg, Softwarepark 32a, 4232 Hagenberg, Austria