Deep Feature-Driven TAO Tree Fusion Model for Autonomous Intrusion Detection in Software-Defined 6G Network Slicing Architectures

Abstract

The evolution of 6G networks introduces significant demands for secure, adaptive, and high-performance 
communication systems. Conventional network monitoring and intrusion detection approaches, largely 
based on static rules and manual inspection, are insufficient for managing the complexity and scale of 
modern network environments. These methods often fail to detect advanced cyber threats and are 
ineffective in optimizing network performance under dynamic conditions. This research proposes a 
Machine Learning (ML)-driven framework based on Classification and Regression Tree (CART) principles 
for dual-purpose analysis, including attack classification and throughput prediction. The system integrates 
multiple ML techniques, such as Support Vector Machine (SVM), k-Nearest Neighbors (KNN), and a 
custom-designed Tree-based Adaptive Optimization (TAO) and an enhanced hybrid Deep Convolutional 
Neural Network (CNN) with TAO-CART (Deep CNN-TAO CART) architecture. The Deep CNN module 
performs automated feature extraction, while the TAO-CART ensemble improves classification and 
regression performance for complex network traffic patterns. These models are trained to accurately 
identify malicious traffic patterns while simultaneously predicting network throughput. To improve model 
performance, preprocessing techniques including Label Encoding, feature scaling through Standardization, 
and class balancing using Synthetic Minority Over-sampling Technique (SMOTE) are applied. This 
integrated approach enhances generalization capability and robustness across diverse network scenarios. 
Experimental results indicate that the Deep CNN – TAO CART ensemble model achieves superior 
classification accuracy along with reliable regression performance compared to individual models. 
Furthermore, the system is implemented through a web-based interface using the Flask framework, enabling 
real-time analysis and user interaction. The proposed solution provides a scalable and intelligent framework 
for improving network security and optimizing performance in next-generation communication systems.