Dual-Task Learning Framework for Next-Gen Network Security and Performance Optimization Using Machine Learning

Abstract

The evolution of 6G networks introduces new challenges for ensuring secure, adaptive, and highperformance communication systems. Traditional network monitoring and intrusion detection approaches, which rely on static rules and manual inspection, are inadequate for handling the complexity and scale of modern network environments. These methods often fail to detect sophisticated cyber threats and are limited in optimizing network performance under dynamic conditions. To address these challenges, this study proposes a machine learning–driven framework based on Classification and Regression Tree (CART) principles for dual-purpose analysis, including attack classification and throughput prediction. The framework integrates multiple machine learning models such as Support Vector Machine (SVM), k-Nearest Neighbors (KNN), and a novel Tree-based Adaptive Optimization (TAO) ensemble model inspired by Random Forest (RF). These models are designed to identify malicious traffic patterns while simultaneously predicting network throughput. To enhance performance and generalization, the system incorporates preprocessing techniques such as Label Encoding, feature standardization, and class balancing using the Synthetic Minority Over-sampling Technique (SMOTE). This ensures robustness across diverse and complex network scenarios. Experimental results demonstrate that the TAO ensemble model outperforms individual models in both classification accuracy and regression reliability. Additionally, the framework is deployed through a web-based interface using Flask, enabling real-time monitoring and user interaction. The proposed system offers a scalable, intelligent, and efficient solution for strengthening cybersecurity and optimizing performance in next-generation communication networks.