2SPUF: Machine Learning Attack Resistant SRAM PUF

Abstract

Internet of Things (IoT) has grown up as an essential aspect of the modern age because it provides comfort to human life by massive connectivity of devices with greater flexibility and control. Security components in IoT systems are very crucial because the devices within the IoT system are exposed to numerous malicious attacks. Typical security components in IoT system performs authentication, authorization, message, and content integrity check. Since IoT systems are resource constraints, it becomes a bit difficult to implement traditional security mechanisms and protocols. For example, authentication is implemented using crypto module, but it is infeasible in IoT domain due to the distributed nature of IoT systems. Physical Unclonable Function (PUF) is considered to be a unique identification of a device that can not be cloned. Hence, PUFs are beneficial in IoT domain to perform basic security operations like authentication, key generation etc. However, there are some attacks proposed on various PUFs using machine learning techniques that model the challenge-response behavior. In this paper, we propose a Two Round SRAM PUF (2SPUF), which shows better resistance to machine learning modeling attacks (ML-MA). We use some well-known machine learning techniques to test ML-MA resistance of 2SPUF design. The result shows that the proposed PUF architecture has better resistance to machine learning modeling attacks.