LSD (LinkedIn Scam Detector) is a project that detects the likelihood of scams or fraudulent behavior on LinkedIn profiles.
It integrates machine learning, web scraping, and session management into a backend-compatible engine.
- Machine Learning Engine: Trained using scikit-learn (Random Forest classifier) to classify LinkedIn profiles as scam or genuine.
- Data Preprocessing: Cleans and structures scraped LinkedIn data for ML model input.
- Automated Scraping: Uses Playwright to fetch structured profile information while handling LinkedIn’s dynamic content.
- Cookie/Session Management: Generates and manages authenticated cookies for scraping.
- Model Persistence & Evaluation: Models are saved with
jobliband evaluated with metrics liker2_score. - Backend Integration: Outputs predictions in JSON format, consumed by a Node.js backend via
child_process.spawn.
- Python: Core logic, ML pipeline, and scraping automation.
- scikit-learn: Random Forest for prediction and model evaluation.
- pandas: Data preprocessing and analysis.
- joblib: Model persistence.
- Playwright (Python): LinkedIn scraping.
- Node.js: Backend service communicating with the Python engine.
- JavaScript (Frontend): For the website interface (built by collaborators).
- Generate cookies for LinkedIn login and place them in
cookies/. - Run the scraper to fetch data from a target profile.
- Preprocess and feed the data into the ML engine.
- Backend consumes the JSON output for integration with the frontend.
- ML and Scraping: Snow Sadh
- Website and Extension Development: Sneha Jha