How lyhenglong/phishing-detector works
Phishing URL Detector extracts structural signals from URLs and classifies them with a trained model. The main user workflow is entering one or more URLs through the Streamlit interface or REST API; the predictor extracts features, scales them, and returns a label, probability, and risk level. A separate training workflow builds features from labeled CSV data, trains and selects a classifier, and saves model artifacts. The README documents evaluation and feature-exploration capabilities; source excerpts show the standalone evaluator loading saved artifacts. Edges are limited to relationships shown in excerpts, the source index, or README; unsampled implementation details are not inferred.
Main entry points: predict.py Prediction service · train_model.py Model training pipeline · evaluate.py Standalone evaluator · build_features.py Feature dataset builder
Components 14
User interfaces
- Streamlit app
app/streamlit_app.py - REST API
app/api.py
Prediction runtime
- Prediction service
src/predict.py - URL feature extractor
src/feature_extraction.py - Trained classifier
models/best_model.pkl - Feature scaler
models/scaler.pkl - Model metadata
models/metadata.json
Training and data
- Labeled URL dataset
data/phishing_dataset_v1.csv - Feature dataset builder
utils/build_features.py - Extracted feature dataset
data/phishing_dataset_v1_features.csv - Model training pipeline
src/train_model.py - Synthetic data generator
utils/generate_dataset.py
Evaluation
- Standalone evaluator
src/evaluate.py
Other components
- URL analyst
Connections 20
- URL analyst → Streamlit app: submits URLs
README.md - URL analyst → REST API: submits URLs
README.md - Streamlit app → Prediction service: requests predictions
app/streamlit_app.py - REST API → Prediction service: requests predictions
app/api.py - Prediction service → URL feature extractor: extracts features
src/predict.py - Prediction service → Feature scaler: loads and applies
src/predict.py - Prediction service → Trained classifier: loads and predicts
src/predict.py - Prediction service → Model metadata: loads metadata
src/predict.py - Labeled URL dataset → Feature dataset builder: provides labeled URLs
README.md - Feature dataset builder → URL feature extractor: extracts features
utils/build_features.py - Feature dataset builder → Extracted feature dataset: writes features
utils/build_features.py - Model training pipeline → Feature dataset builder: builds if missing
src/train_model.py - Model training pipeline → Extracted feature dataset: loads features
README.md - Model training pipeline → Trained classifier: saves selected model
README.md - Model training pipeline → Feature scaler: saves scaler
README.md - Model training pipeline → Model metadata: saves metadata
README.md - Standalone evaluator → Trained classifier: loads model
src/evaluate.py - Standalone evaluator → Feature scaler: loads scaler
src/evaluate.py - Standalone evaluator → Model metadata: loads and updates
src/evaluate.py - Standalone evaluator → Extracted feature dataset: reads evaluation data