Cybersecurity Intrusion Detection with Hilbert Curve

Published at VISAPP 2024 · Rome, Italy

Problem

Traditional tabular IDS features can miss spatial patterns. This project transforms flow features into image-like representations for CV models.

Pipeline

• Preprocess network flow logs.
• Apply Hilbert Curve mapping to convert features into 2D image matrices.
• Train CNN-based classifier for intrusion type prediction.
• Deploy SOC-ready inference workflow.

Training Snapshot

Tech Stack

Python, TensorFlow, scikit-learn, Jupyter, Azure compute resources.