ItsMakion/malawi-smishing-detector

A lightweight SMS/ Whatsapp smishing detector tailored for Malawi # Malawi Smishing Detector A lightweight SMS/WhatsApp smishing detector tailored for Malawi's mobile money ecosystem (Airtel Money, TNM Mpamba). Built with Python, scikit-learn, and rule-based detection. Includes a small demo bot and sample dataset. --- ## 🚨 Why this project? Mobile money scams are one of the biggest cyber threats in Malawi. Attackers send fake SMS or WhatsApp messages like: - "Verify your Airtel PIN to keep your account active" - "TNM Mpamba: You have received K2,500. Click link to claim" - Messages in Chichewa tricking users into sending codes or money This tool shows how **local-language smishing detection** can work in practice. --- ## 🛠 How it works 1. **Dataset** - `data/sample_messages.csv` → example SMS/WhatsApp messages (English & Chichewa). - `data/labels.csv` → whether each is `ham` (safe) or `smishing`. 2. **Detection methods** - **Rule-based detector**: Looks for suspicious keywords/phrases (e.g. “verify PIN”, “Mpamba”, “agent”). - **Machine learning model**: Uses scikit-learn (TF-IDF + Logistic Regression) to classify new messages. 3. **Demo bot** - Simple Telegram bot that lets you paste a message. - It replies with `Likely smishing` or `Looks safe`. 4. **Notebook demo** - `demo/notebook_demo.ipynb` → shows how to load data, train model, and test rules. --- ## 📊 Features - Keyword-based detection with local language awareness (English + Chichewa). - Lightweight ML model (trainable in seconds). - Telegram bot demo for real-time message checks. - Documented rules in `docs/RULES.md`. - Reporting flow (`docs/REPORTING_FLOW.md`) → how a detected scam could be reported to **mwCERT**. --- ## 🔧 Installation ```bash git clone github.com cd malawi-smishing-detector