An Intelligent Digital Twin Model for Attack Detection in Zero-Touch 6G Networks

The rapid adoption of Internet of Things (IoT) services and the increasingly stringent dependability and performance requirements are transforming next-generation wireless network management towards zero-touch 6G networks. Zero-touch management is also particularly attractive in the context of smart industrial production (Industry 5.0) as it allows more autonomous control over all operations without human intervention. However, such a paradigm requires real-time remote monitoring and control throughout the life cycle of all physical entities. Against this backdrop, digital twin is a critical technology for Industry 5.0 and beyond for novel network automation solutions. A digital twin enables fine-grained monitoring, testing and experimentation by providing a model of its physical counterparts. In this regard, we propose a three-layered digital twin framework's architecture to effectively bridge the gap between the physical world and the digital world in zero-touch 6G networks. As a case study, we consider an intelligent digital twin model to detect cyber attacks, such as brute force, web attacks, and DDoS attacks. However, an excessive increase in data volume requires exploratory efforts for efficient and scalable attack detection. In addition, due to the insufficient annotation and imbalanced distribution of classes, challenges are introduced regarding the achievable accuracy. To address these issues, we propose the use of Long Short-Term Memory Network (LSTM) with the combination of the Synthetic Minority Over-sampling Technique (SMOTE) for attack detection for the new and unseen samples with the assistance of digital twin. We also analyze samples within the cluster and eliminate samples that are somewhere between the clusters and negatively affect the final accuracy by silhouette score. Extensive experiments show that our proposed intelligent digital twin model for attack detection improves the accuracy performance.