Neural IDS: Real-Time DDoS Detection

Neural IDS is a deep learning intrusion detection system trained on the CICIDS DDoS subset, using a hybrid CNN-LSTM model that processes 10 time steps of 78 features per input. A Flask API serves predictions, Prometheus scrapes the metrics, and Grafana turns them into a real-time dashboard you can actually watch during an attack.

Built and trained for educational and research purposes not a production security tool.

Why: Wanted experience working with real static, labelled data training against CICIDS instead of a toy dataset. Grafana can alert straight to Telegram when a prediction crosses a threshold, so wiring that up became part of it too.

Stack: Python, TensorFlow/Keras, Flask, Prometheus, Grafana