Countering PUF Modeling Attacks through Adversarial Machine Learning

Abstract

A Physically Unclonable Function (PUF) is an effective option for device authentication, especially for IoT frameworks with resource-constrained devices. However, PUFs are vulnerable to modeling attacks which build a PUF model using a small subset of its Challenge-Response Pairs (CRPs). We propose an effective countermeasure against such an attack by employing adversarial machine learning techniques that introduce errors (poison) to the adversary's model. The approach intermittently provides wrong response for the fed challenges. Coordination among the communicating parties is pursued to prevent the poisoned CRPs from causing the device authentication to fail. The experimental results extracted for a PUF implemented on FPGA demonstrate the efficacy of the proposed approach in thwarting modeling attacks. We also discuss the resiliency of the proposed scheme against impersonation and Sybil attacks.

Entropy Sources Based on Silicon Chips: True Random Number Generator and Physical Unclonable Function

“… Contaminate the data accessible by the attackers. There are some design ideas that poison the response data collected by the attackers, and the legitimate user knows how to differentiate the true and fake responses [ 107 ]. In [ 108 ], by adding some extra models, such as PRNG and a fake PUF, an active deception protocol was guaranteed to prevent ML attacks.…”

SWeeT: Security Protocol for Wearables Embedded Devices’ Data Transmission

“…Thanks to being lightweight and constituting a unique electronic signature, Physically Unclonable Functions (PUFs) is deemed a promising means for generating secret keys. However, PUF-based protocols can be vulnerable to contemporary attacks such as behavior modeling, message replay, and impersonation [13]. To mitigate vulnerability to these attacks, Chatterjee et al [14] have proposed using a combination of PUFs and asymmetric cryptographic modules.…”