AI/ML-Assisted Threat Detection and Mitigation in 6G Networks with Digital Twins: The HORSE Approach
Fabrizio Granelli, Malak Qaisi, Panagiotis Kapsalis, Panagiotis K. Gkonis, Νικόλαος Νομικός, Iulisloi Zacarias, Admela Jukan, Panagiotis T. Trakadas · 2024
In this paper a novel architectural approach is presented for threat detection and mitigation via machine learning (ML) in sixth-generation (6G) networks. In this context, due to the vast number and polymorphic nature of potential attacks, novel approaches are required that are based on a holistic data monitoring and evaluation context. Therefore, the HORSE (holistic, omnipresent, resilient services) framework is described, where the concept of digital twins is integrated with ML techniques for proper threat detection and mitigation. To this end, the main HORSE components are analyzed, including ML and deep learning (DL) training for threat detection as well as efficient ML function orchestration and intent-based networking to apply the mitigation actions and reconfigure the network if necessary. Moreover, two indicative use cases are presented and analyzed. The first one is related to threat prediction, while the second one to threat mitigation. In both cases, the sequence of actions is described along with the corresponding components that formulate a closed-loop network reconfiguration and optimization process.