Fortified Edge 3.0: A Lightweight Machine Learning based Approach for Security in Collaborative Edge Computing

Machine Learning (ML) at the edge has the advantages of lower costs, reduced bandwidth needs, easy access, and high security. Machine Learning-based security protocols are hence suitable for providing added security at the edge, especially in distributed environments like collaborative edge computing (CEC). Security models are developed using the different ML methods, supervised, unsupervised, and reinforced learning. This research explores the ML methods suitable for security at the edge, particularly the supervised ML methods. The purpose of this research is to test the various supervised ML methods and compare their metrics to propose an efficient method for providing security at the edge. ML at the Edge can help tackle security challenges like external attacks, anomaly detection, data security privacy, and trust management. The edge environment considered is the Edge Data Centers (EDCs) in CEC, the security feature is the Physical Unclonable Function Certificate Authority (PUF CA) based authentication system, which is designed for authenticating EDCs during dynamic load balancing and the ML model is used to monitor the authentication process for intrusion and anomaly detection.