A Machine Learning Attack-Resistant PUF-Based Robust and Efficient Mutual Authentication Scheme in Fog-Enabled IoT Environments

Fog-enabled Internet of Things (IoT) systems have a lot of attention and are being applied in various fields, including smart homes, smart healthcare, smart factories, and smart grids. These fog-enabled IoT systems enhance citizens’ quality of life and provide innovative and high-quality IoT services. However, these systems can be vulnerable to cyber security attacks since an adversary attempts to modify, delete, block, and intercept the exchanged data over an insecure channel. Besides cyber security attacks, IoT can be fragile to physical security attacks because they are deployed in hostile environments. Physical unclonable function (PUF) is a promising solution to address these issues. PUF can protect the security of IoT devices with minimal computation costs against cyber/physical attacks from an adversary. However, with recent advances in artificial intelligence (AI) technology, existing PUFs used in authentication and key agreement (AKA) schemes are susceptible to machine-learning (ML)-based modeling attacks. To address these challenges, we design the ML-based modeling attack-resistant PUF-based robust and efficient AKA scheme in fog-enabled IoT environments. We evaluate the security of the proposed scheme by performing informal and formal security analyses, such as ROR oracle model and AVISPA simulation. We present the implementation to demonstrate the accuracy against ML-based modeling attacks. Moreover, we perform the performance comparison analysis between the proposed scheme and existing schemes based on testbed implementation. Consequently, the proposed scheme provides superior security and efficiency compared to existing schemes and can be suitable for practical fog-enabled IoT systems.