A lightweight FastText-based machine learning model to classify domain names as phishing or clean in real time. Trained on a dataset of 2 million labeled domains using Facebook Research's FastText library with n-gram support for improved pattern recognition. Offers both command-line prediction and a containerized REST API via Docker. Returns probability scores for each classification, making it suitable for integration into security pipelines, email filters, and DNS monitoring systems. Fast inference with minimal resource usage.
git clone https://github.com/mstfknn/phishing-fasttext-model.git
# Command-line prediction
echo "carreeffoursa.site" | ./fasttext predict phishing_model.bin -
# Docker API deployment
docker run -p 8080:8080 mstfknn/phishing-fasttext:latest
curl -X POST http://localhost:8080/predict \
-H "Content-Type: application/json" \
-d '{"domain": "carreeffoursa.site"}'