chaithanyabs441-ux/phishing-website-detector

How chaithanyabs441-ux/phishing-website-detector works

PhishReaper is a phishing-URL triage service backed by a reproducible classical-ML research pipeline. The deployed flow accepts a URL and optional uploaded HTML, extracts URL/HTML features, scores with a calibrated CatBoost model or URL-SVM fallback, and can override the verdict on a Safe Browsing threat match. A separate offline workflow downloads data, creates leakage-controlled splits, builds features, compares models, and calibrates thresholds. Serving relationships are supported by the app excerpt and SOURCE INDEX; broader research-stage links follow the README, since several evaluation modules were not sampled.

Main entry points: app.py Scan API · phase4_boosting.py Tree tuning · url_features.py URL features · phase4_data.py Training data

Components 20

Online triage

Scan API webapp/app.py
Request guards src/serving/request_security.py
Safe HTML fetch src/serving/safe_fetch.py
Model registry src/serving/model_registry.py
Triage verdict webapp/app.py

Feature extraction

URL features src/features/url_features.py
HTML features src/features/html_features_v2.py

Model research

Dataset download src/data/download_dataset.py
Blocked splits src/evaluation/create_blocked_splits.py
HTML feature cache src/evaluation/build_html_features_v2_cache.py
Training data src/evaluation/phase4_data.py
SVM baselines src/evaluation/evaluate_svm_blocked.py
Tree tuning src/evaluation/phase4_boosting.py
Text models src/evaluation/phase5_text.py
OOF stacking src/evaluation/phase6_stacking.py
Calibration src/evaluation/phase7_calibration.py
Benchmark evaluation src/evaluation/evaluate_url_html_official.py

Threat intelligence

Safe Browsing lookup src/reputation/google_safe_browsing.py

Other components

Analyst
Google Safe Browsing

Connections 23

Analyst → Scan API: submits scan README.md
Scan API → Request guards: validates request webapp/app.py
Scan API → Safe HTML fetch: optionally fetches webapp/app.py
Scan API → URL features: extracts features webapp/app.py
Scan API → HTML features: parses HTML webapp/app.py
Scan API → Model registry: requests prediction webapp/app.py
Model registry → Triage verdict: returns model score README.md
Scan API → Safe Browsing lookup: checks URL webapp/app.py
Safe Browsing lookup → Google Safe Browsing: queries threat list src/reputation/google_safe_browsing.py
Scan API → Triage verdict: renders triage README.md
Triage verdict → Analyst: presents signal README.md
Dataset download → Blocked splits: provides shards README.md
Blocked splits → SVM baselines: defines holdout README.md
HTML feature cache → Training data: supplies HTML cache src/evaluation/phase4_data.py
URL features → Training data: computes URL inputs src/evaluation/phase4_data.py
Training data → Tree tuning: loads model data src/evaluation/phase4_data.py
SVM baselines → Text models: baseline comparison README.md
Tree tuning → Text models: model comparison README.md
Text models → OOF stacking: provides stack inputs README.md
OOF stacking → Calibration: provides OOF scores README.md
Tree tuning → Calibration: selects finalist README.md
Benchmark evaluation → URL features: uses shared features src/evaluation/evaluate_url_html_official.py
Benchmark evaluation → HTML features: uses HTML features src/evaluation/evaluate_url_html_official.py

More Python diagrams: lalitgehani/snackbase · nayakwadis/mftool · onemoh/html-explainer · openvax/mhcflurry · somnusochi/vlm-autoyolo

Drawn by GitDiagram on November 12, 2026 from the repository's file tree, README and source files. Read as Markdown ·