SmartNet AI: A Unified Framework for Cyber Threat Detection and Network Performance Optimization

Abstract

The advancement of 6G networks brings increasing requirements for secure, flexible, and high-efficiency communication infrastructures. Traditional network monitoring and intrusion detection methods, which rely heavily on fixed rules and manual analysis, are no longer adequate for handling the growing complexity and scale of modern network systems. These approaches often struggle to identify sophisticated cyber threats and fail to maintain optimal performance in highly dynamic environments. This study introduces a Machine Learning (ML)-based framework grounded in Classification and Regression Tree (CART) methodology, designed for dual-function analysis encompassing both attack detection and throughput estimation. The proposed system incorporates various ML algorithms, including Support Vector Machine (SVM), k-Nearest Neighbors (KNN), and a novel Tree-based Adaptive Optimization (TAO) ensemble model influenced by Random Forest (RF) techniques. These models are trained to effectively detect malicious network behavior while also forecasting network throughput. To enhance model efficiency, several preprocessing strategies are employed, such as label encoding, feature scaling through standardization, and dataset balancing using the Synthetic Minority Over-sampling Technique (SMOTE). These steps contribute to improved robustness and generalization across varying network conditions. Experimental evaluations demonstrate that the TAO ensemble model outperforms individual models in terms of classification accuracy while maintaining strong regression capabilities. Additionally, the system is deployed through a web-based platform using the Flask framework, allowing real-time analysis and user interaction. Overall, the proposed framework offers a scalable and intelligent solution for strengthening network security and improving performance in next-generation communication environments.