Integration Of Real-Time NetFlow Streaming Analytics For Adaptive Network Attack Detection

Authors: Pavithra V, Dr. D. Rajinigirinath, T. Saranya

Abstract: In modern digital infrastructures, network intrusion detection systems (NIDS) require real-time capabilities to effectively identify and prevent ongoing cyberattacks. This paper presents the integration of real-time NetFlow streaming analytics into an adaptive AI-based attack detection framework. Traditional machine learning approaches are often limited to offline datasets, making them less effective for live monitoring. The proposed system employs streaming NetFlow data collection, continuous preprocessing, and online inference using Random Forest and Support Vector Machine models. A real-time analytics dashboard built using Streamlit provides live alerts and visualization of suspicious flows. Experimental results indicate a detection latency reduction of 35% and a 10% improvement in adaptive accuracy when using continuous learning updates. This work demonstrates that integrating real-time NetFlow analytics significantly enhances the responsiveness and scalability of intrusion detection systems.

DOI: http://doi.org/10.5281/zenodo.17471011