AI-powered fraud & phishing detection on WhatsApp.
Problem & solution
WhatsApp fraud grows every year, and most victims have no tools to verify whether a message is legitimate. SecurityBot – WA solves that problem: a system that analyzes messages, links and images directly inside WhatsApp to detect phishing and fraud, combining a machine learning model (SVM) with contextual AI analysis. The system retrains continuously through an RLHF loop powered by real user feedback, and runs on a serverless Google Cloud architecture that scales automatically with demand.
Highlights
- >90% accuracy
- Serverless on Google Cloud
- RLHF
- Anti-fraud OCR
What it does
- [01]
FastAPI architecture deployed on Google Cloud Run and integrated with Cloud SQL (PostgreSQL) — serverless and scalable.
- [02]
Hybrid phishing detection with SVM (scikit-learn) + contextual AI analysis, achieving over 90% accuracy.
- [03]
RLHF (Reinforcement Learning from Human Feedback) pipeline for iterative model retraining.
- [04]
Automated CI/CD with Cloud Build + Docker.
- [05]
OCR (Tesseract + OpenCV) to analyze fraudulent images and detect manipulated screenshots.
Try the app
Demo video coming soon
This project is a WhatsApp bot, so it can't be embedded. While the demo video is on its way, check out the screenshots and architecture below.
Screenshots
Click a screenshot to enlarge it
Tech stack
- Python
- FastAPI
- PostgreSQL
- Google Cloud Run
- Cloud SQL
- Docker
- scikit-learn
- Tesseract
- OpenCV
- CI/CD