Lightweight ML-Based Network Anomaly Detection System | Cyber AI Hackathon 2025
This video demonstrates my Lightweight Machine-Learning Based Network Anomaly Detection System , developed for the Cyber AI Hackathon 2025.
"Intrusion creates Anomaly"
Traditional IDS tools rely on large signature
databases and struggle to detect unknown or zero-day attacks. This project takes a different approach by using unsupervised machine learning to learn normal network behavior and flag anomalies in real time.
💡 Key Features:
• Real-time packet capture and flow aggregation
• Behavior-based anomaly detection using Isolation Forest
• Threat classification (Port Scan, DoS-like, ICMP anomalies)
• Live dashboard with score history, protocol distribution, alerts, and traffic stats
• Fully lightweight -runs on a laptop or Raspberry Pi
• Zero signatures required -detects unknown attacks instantly
🛠 Tech Stack:
Python, Scapy, Flask, Scikit-Learn, JS Charts
🎯 End Users:
Small labs, home networks, IoT environments, college setups, or anyone without access to enterprise-grade security tools.
This project was designed and implemented end-to-end by Utkarsh Mishra for team Mishaz (Team T-2974).
Thanks for watching!