Scam Protect

Screen once a scam has been detected
All the contacts that will be emailed once a likely scam has been detected (via email or phone depending on urgency)
Alert page to review any alerts you have received
Home page. Turn off scamshield to disable all features
Overlay once a high risk app has been detected to alert the user that risk monitoring has been activated

Inspiration

A close friend's grandmother was scammed out of thousands. This made us realize that more than $10 billion is taken out as scams annually. We wanted to fix this

What it does

When it detects a scam, it alerts the user and emails their emergency contacts instantly. ScamProtect runs silently in the background, and starts monitoring screen activity, apps, and audio in real time if it detects something suspicious according to our algorithm and agentic AI.

How we built it

Electron frontend, Python/FastAPI backend. A two-stage Gemini AI pipeline analyzes screen and audio every 30 seconds — a fast model scores risk, a heavier model confirms and summarizes. Alerts stream via SSE and trigger email notifications.

Challenges we ran into

Electron's context isolation broke our real-time streaming. API rate limits forced us to build a fallback system. Stale state persistence caused phantom alerts.

Accomplishments that we're proud of

A fully working real-time scam detection system — visual, audio, app tracking, full-screen alerts, and automatic email notifications.

What we learned

Real-time desktop monitoring is deceptively complex. We learned to navigate Electron's security model, orchestrate multi-stage AI pipelines, and understand how scammers actually operate.