Terahertz time-domain spectroscopy (THz-TDS) fingerprinting for integrated circuit (IC) identification in tracking and tracing applications

The global outsourcing of semiconductor fabrication has led to hardware security concerns such as counterfeit integrated circuits (ICs) and Hardware Trojans (HTs), compromising the trustworthiness of semiconductor devices in critical applications. To address the issue of counterfeit ICs and HTs, various physical inspection methods have been developed. These methods, which include X-ray imaging, Scanning Acoustic Microscopy (SAM), and Scanning Electron Microscopy (SEM), are employed to detect irregularities within the packaging of ICs, aiding in the identification of counterfeit samples and the detection of HTs. Previous studies have shown that encapsulant material differences in counterfeit ICs can be detected by observing the refractive index variance between genuine and counterfeit products. This is achieved by measuring layer thickness and time delay in THz-TDS. THz-TDS employs a pulsed Terahertz signal to discern the effective refractive index differences between authentic and counterfeit IC packaging. However, anomaly detection often requires high resolution, which is time-consuming and necessitates standard samples for comparison, which are challenging to obtain. In this research, we focus on generating a THz-TDS 'fingerprint' for each IC sample for hardware assurance, rather than detecting packaging anomalies. This paper explores using both supervised and unsupervised machine learning models to demonstrate the effectiveness of THz-TDS 'fingerprinting' in IC sample identification. We also investigate the tolerance of THz-TDS data collection locations to identify various types of IC packaging. This involves collecting THz-TDS data from different IC packaging samples at multiple locations to assess the impact on accuracy in sample identification.