In the realm of Internet of Things (IoT) security, the Physical
Unclonable Function (PUF) emerges as a viable solution for generating
individualized secure keys. These keys play a vital role in
authentication, chip protection, and supply chain security, serving as
crucial safeguards for resource-constrained IoT devices. This work
introduces Boosted Configurable Ring Oscillator PUF (BC-PUF), an
area-efficient variant of CROPUF that suits resource-constrained IoT
devices. Additionally, BC-PUF employs an absorbent approach for the
intermediate responses to optimize the utilization of CMOS delay
configurations. Moreover, BC-PUF mitigates the risk of Machine Learning
(ML)-based modeling attacks. BC-PUF is a multi-bit response architecture
that has been developed, analyzed, and evaluated across 50 FPGAs, from
each dataset of 10K Challenge-Response Pairs (CRPs) is extracted.
Experimental results report average values of 50.2%, 40.4%, 42%,
3.925%, and 83% for uniformity, diffuseness, uniqueness, reliability,
and Shannon entropy, respectively. In comparison to state-of-the-art
designs, the BC-PUF design achieves area reductions ranging from 15% to
96.7% across various architectures and implementations, along with a
power reduction of 9.7%. Moreover, BC-PUF demonstrates resilience
against ML-based attacks, achieving a prediction accuracy of 51%.