SFGAD is a tool for detecting anomalies in graph and graph streams with python.

I provides:

• Efficient computation of graph features
• Statistical models for detecting anomalous behavior
• Graph scanning to detect connected graph anomalies
• A customizable detection framework with 6 components
• Several pre-defined configurations

Process

---

Installation

---

Dependencies

• Python: 3.5 or higher
• NumPy: 1.8.2 or higher
• SciPy: 0.13.3 or higher
• Pandas: 0.22.0 or higher
• NetworkX: 1.11.0 or higher

Installation

Installation of the latest release is available at the Python package index.

pip install sfgad

The source code is currently available on GitHub: https://github.com/sudrich/sf-gad

Testing

For testing use pytest from the source directory:

pytest sfgad

Usage

The framework defines an modular interface that allows full customization of the analysis process. For examples, see the tutorials on using a pre-defined analyzer and using a custom analyzer.

Acknowledgements

This work originated from the QuestMiner project (grant no. 01IS12051) and was partially funded by the German Federal Ministry of Education and Research (BMBF).