ML-Defense: Machine Learning for building Dependable Federated Network System

Federated network platforms over the 5G and be-yond networks will extend novel opportunities to provide end-to-end Internet of Things (IoT) services. Using a logically centralized software-defined controller (i.e., SDN) is more practical than conventional distributed and ad hoc management to support efficient, scalable management and orchestration. However, the security issues of IoT services over the federated platforms using SDN are not well explored. A distributed wormhole attack over the federated IoT system is one of the most challenging security issues, where a group of secret attackers can easily exploit the centralized controllers with various wireless control messages. This paper presents an ML-Defense (Machine Learning for building a Dependable Federated Network System) against various intelligent wormhole attacks. Using an unsupervised learning algorithm, ML-Defense identifies different wormhole attacker patterns and locations. Our simulation results indicate that ML-Defense can cluster the attackers' location without requiring special devices and pinpoints the core of the attackers by examining the suspicious area.