SecureFlow Analyzer: Live Packet Inspection and Threat Detection Framework

Abstract

The rapid growth of digital communication and network-based systems has significantly increased network traffic, making monitoring and security more challenging. Traditional network analysis methods, which rely on manual inspection and rule-based systems, are effective for detecting known patterns but struggle with evolving and complex threats. These approaches lack adaptability, are timeconsuming, and are inefficient when handling large-scale data. A key challenge is accurately classifying network activities and detecting anomalies in dynamic environments while addressing issues such as data imbalance, noise, and high dimensionality. To overcome these limitations, this research proposes a machine learning-based framework designed to improve network traffic analysis. The framework integrates data preprocessing, exploratory data analysis, and class balancing using the Synthetic Minority Over-sampling Technique (SMOTE). It employs multiple classification models, including Ridge Classifier (RC), Linear Discriminant Analysis (LDA), and Decision Tree (DT) as baseline methods. Additionally, a hybrid Neuro Tree Fusion (NTF) model is introduced, combining Multi-Layer Perceptron (MLP) and Decision Tree (DT) to enhance classification performance. The proposed approach improves accuracy, adaptability, and robustness by leveraging both statistical and non-linear learning techniques. A graphical user interface (GUI) is also developed to facilitate user interaction, allowing dataset upload, analysis, and prediction generation. Experimental results demonstrate significant performance improvement, with the hybrid model achieving an accuracy of 99.13%, outperforming all baseline models. This study offers a scalable, reliable, and high-performance solution for modern network traffic classification and security analysis.