Abstract
In the recent years, Physical Unclonable Functions(PUFs) are emerged to be one of the lightweight hardware security primitives for device authentication, identification, such as Internet of things (IoT). IoT comprises connection of multiple number of nodes (devices) for exchanging the information across different networks. PUFs can sense the minute and unavoidable process variations during the fabrication process and generates the unique number of challenge-response pairs(CRPs), which can be stored and extensively used for secure associations between smart devices in IoT. Arbiter PUFs and ring oscillator PUFs are most commonly used strong PUFs in current day scenario. The conventional Linear arbiter PUFs are suffers from low reliability and vulnerable to Machine Learning attacks. In this paper, we proposed a Current starved Inverter (CSI) based arbiter PUF which enhances the non- linearity and randomness. The Proposed architecture was simulated using cadence spectre CMOS 45nm technology and estimated its metrics such as uniqueness reliability and uniformity.