Robust Network Anomaly Detection with K-Nearest Neighbors (KNN) Enhanced Digital Twins
Peprah Obed Adjei, Sumit Kumar Tetarave, Caroline Sangeetha John, Madlyn Manneh, Parthasarathi Pattnayak · 2024
Modern network security remains a critical concern in the digital landscape due to evolving cyber threats and increasingly sophisticated attack vectors such as Advanced Persistent Threats and Zero-Day Vulnerabilities. Leveraging advanced tech-nologies such as artificial intelligence (AI) and machine learning (ML) can enhance threat detection capabilities and improve incident response times when detecting and mitigating network security threats. On the other hand, an imbalanced dataset of network traffic in AI/ML models presents several challenges and can significantly impact the performance and effectiveness of the models to predict attacks. Our research aims to amplify the robustness of the imbalanced network traffic dataset to fit the analysis and adaptability of KNN-based Digital Twins dedicated to network anomaly detection. This paper capitalizes on the remarkable performance of the model, characterized by impeccable precision, recall, and F1-score, as indicated by the classification report with 99% accuracy. The confusion matrix further highlights the model's performance using the proposed robustness dataset, showing a minimal False Positive Rate (FPR) compared to similar works in the literature.