Abstract
Authentication is a crucial security service for the wireless sensor networks (WSNs) in versatile domains. The deployment of WSN devices in the untrusted open environment and the resource-constrained nature make the on-chip authentication an open challenge. The strong physical unclonable function (PUF) came in handy as light-weight authentication security primitive. In this paper, we present the first ring oscillator (RO) based strong physical unclonable function (PUF) with high resilience to both the electromagnetic (EM) side-channel attack and the support vector machine (SVM) modelling attack. By employing an RO based PUF architecture with the current starved inverter as the delay cell, the oscillation power is significantly reduced to minimize the emitted EM signal, leading to greatly enhanced immunity to the EM side-channel analysis attack. In addition, featuring superior reconfigurability due to the conspicuously simplified circuitries, the proposed implementation is capable of withstanding the SVM modelling attack by generating and comparing a large number of RO frequency pairs. The reported experimental results validate the prototype of a 9-stage RO PUF fabricated using standard 65 nm complementary-metal-oxide-semiconductor (CMOS) process. Operating at the supply voltage of 1.2 V and the frequency of 100 KHz, the fabricated RO PUF occupies a compact silicon area of 250 m and consumes a power as low as 5.16 W per challenge-response pair (CRP). Furthermore, the uniqueness and the worst-case reliability are measured to be 50.17% and 98.30% for the working temperature range of −40∼120 C and the supply voltage variation of ±2%, respectively. Thus, the proposed PUF is applicable for the low power, low cost and secure WSN communications.
Highlights
Wireless sensor networks (WSNs) are intensely ubiquitous and encompass a broad range of new applications, such as AR/VR, Internet of Things (IoT), and vehicle network [1]
Merli’s work demonstrates that the electromagnetic (EM) measurements are capable of disclosing both the frequency and the location of each ring oscillator (RO), which enables the prediction of the RO physical unclonable function (PUF)’s challenge-response pair (CRP) [12]. To address these security threats, this paper presents the first RO based strong PUF design to resist the EM side-channel attack and the support vector machine (SVM) modelling attack
The setup of the probe station for the post-silicon test is shown in Figure 5, where Agilent oscilloscope with 1GS/s sampling rate is used to measure the output frequencies of the current starved ROs and capture the responses of the PUF
Summary
Wireless sensor networks (WSNs) are intensely ubiquitous and encompass a broad range of new applications, such as AR/VR, Internet of Things (IoT), and vehicle network [1]. The strong PUF is more suitable for the wide range of device authentication applications than its counterpart of weak PUF [9,10,11]. Merli’s work demonstrates that the electromagnetic (EM) measurements are capable of disclosing both the frequency and the location of each RO, which enables the prediction of the RO PUF’s CRPs [12]. To address these security threats, this paper presents the first RO based strong PUF design to resist the EM side-channel attack and the support vector machine (SVM) modelling attack.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call