Inspiration
In East Africa, mobile money has revolutionized how people transact — but it has also opened the door to new fraud risks. Fake SMS alerts, impersonation calls, and threatening messages trick users into losing their hard-earned money. Seeing mobile money agents, students, and small vendors fall victim to these scams inspired us to build a tool that empowers them with instant, accessible protection.
What it does
SilentGuard is an AI-powered fraud protection app that: Detects and flags scam messages in real time. Alerts users before completing suspicious transactions. Monitors for unusual activity like repeated failed PIN attempts. Checks location and transaction timing for risk analysis. Lets users report scams and share verified fraud data. Provides easy-to-understand anti-fraud tips in local languages.
How we built it
We developed a web-based prototype with a mobile-first design using: Frontend: HTML, CSS, Backend: Python (Django). AI & NLP: TensorFlow for scam message classification. Database: PostgreSQL for storing scam patterns and user reports. APIs & Services: Africa’s Talking for SMS handling and mobile money simulation. Tunneling & Testing: Ngrok for exposing the local server to receive webhook callbacks during development. Our AI model was trained on a combination of sample scam messages, public fraud datasets, and locally collected text samples.
Challenges we ran into
Limited access to real, verified local scam datasets. Designing an interface that’s simple for non-tech-savvy users. Simulating mobile money transactions without compromising security. Optimizing for low-bandwidth rural environments.
Accomplishments that we're proud of
Building a functional scam detection prototype in a short hackathon timeframe. Creating a simple, user-friendly design tailored to vulnerable users. Training an initial detection model with promising early accuracy.
What we learned
Fraud patterns vary greatly by region — localization is key. Continuous AI model retraining is critical to stay ahead of new scams. Users trust alerts more when the reason for flagging is clearly explained.
What's next for SilentGuard
Launching a full mobile app with real-time alerts. Partnering with telecoms and fintech companies for stronger fraud prevention. Adding multi-language and voice-based features to reach non-literate users. Expanding to a global scam alert network with live fraud maps.
Log in or sign up for Devpost to join the conversation.