PUF Based Cybersecurity Architecture for Hardware Trojan Detection in IoT Enabled Cyber Physical Systems

In this research, a challenge adaptable Physically Unclonable Function (PUF) is used to offer a novel security architecture for hardware Trojan detection in IoT enabled Cyber Physical Systems (CPS). In the proposed technique, by tracking the fluctuation of entropy and cross correlation between the test answers and a golden reference, a dynamic PUF characterization engine makes it possible to identify minute behavioral abnormalities caused by malevolent circuitry. The system detects Trojans by comparing Pearson’s correlation coefficients and entropy changes with empirically determined thresholds. The design creates multidimensional feature vectors by combining timing behavior with side-channel fingerprinting power. Using golden chip profiles as training data, these vectors are fed into a lightweight machine learning classifier. A federated weighted based classifier that can identify unusual answers that fall outside of the taught decision boundary. This dual mode detection approach combines side channel signatures with statistical signal assessments to increase detection accuracy to 90. 09% and Trojan free accuracy to 95. 8%. Both entropy and correlation based metrics may detect irregularities and are suitable for resource constrained CPS systems, as evaluated in an experiment utilizing ISCAS benchmark circuits.