GENERATIVE AI-DRIVEN NETWORK DIGITAL TWINS: A NOVEL FRAMEWORK FOR MULTIVARIATE ANOMALY DETECTION AND DATA IMPUTATION IN 6G ECOSYSTEMS

Abstract

The symbiotic Internet of Things (S-IoT) networks, powered by 6G technologies, have created and will continue to create a revolution and complexity that necessitates a broader understanding of digital systems. Enabling distributed intelligence through the integration of centralized, peripheral, and cloud computing for automation relies on the support and symbiosis of artificial intelligence. This leads to all integral forms with respect to the largest scale models of the AI and their connection for the resource-limited devices, resulting in serious security vulnerabilities. The root cause is loose endpoints, which are increasingly becoming entry points for sophisticated cyberattacks. Data breaches are becoming increasingly apparent, and system disruptions further exacerbate these vulnerabilities. This study addresses this issue. This framework of the employs adaptive data preprocessing, with addition the  outlier mitigation, class balancing using Adaptive Synthetic Sampling, and minimum-maximum normalization, followed by hierarchical integration and spatial pattern capture dynamics (such as packet length variation and TCP tag anomalies).