begdouri-marwa/Phishing_Analysis_Project

# PhishGuard — Multilingual Phishing Detection System PhishGuard is a Flask-based NLP application designed to detect phishing and suspicious social engineering content in English, French, and Arabic. The project combines dataset retrieval, URL feature analysis, and optional PDF knowledge ingestion into a retrieval-augmented generation (RAG) workflow. **Database spreadsheet:** docs.google.com ## lien de la video drive.google.com --- ## What this project does - Detects phishing and legitimate messages across email, SMS, URL, and social media channels - Uses a hybrid retrieval pipeline with FAISS embeddings and BM25 search - Builds contextual analysis from similar dataset examples - Extracts URL features and suspicious tokens for better phishing detection - Supports PDF knowledge base uploads for extending domain knowledge - Displays dataset dashboard metrics and recent analysis history ## 📁 Repository structure - `app.py` — Flask web server and API endpoints for analysis, PDF upload, and registry management - `requirements.txt` — Python packages required to run the application - `requirements.py` — helper script to inspect and install requirements - `src/` — core backend implementation - `src/rag_pipeline.py` — RAG orchestration, URL detection, and analysis logic - `src/retrieval_module.py` — dataset retrieval system using `faiss`, `sentence-transformers`, and `rank_bm25` - `src/pdf_knowledge.py` — PDF extraction, chunk creation, embedding, and retrieval support - `templates/` — Jinja2 HTML templates for the user interface - `static/` — frontend styles and client-side JavaScript - `data/` — dataset files, indexes, PDF registry, and persistent PDF knowledge artifacts - `uploads/` — uploaded PDF storage - `notebooks/` — experimental notebooks used for RAG research, analysi …