GuardianShield

Last Diwali, my grandfather received a call from someone claiming to be CBI. For three hours, he sat alone on a video call with a man in a fake police uniform, terrified his bank accounts would be frozen. We found him shaking, nearly ready to transfer money to a "safe account." We were lucky—he hadn't sent anything yet.

Then I read about the 67-year-old Hyderabad woman who lost ₹55 million over 17 days of digital house arrest. It could have been my grandfather. It could be anyone's family.

That's when I realized: we're solving the wrong problem.

Current solutions all fail at the same point: they require the victim to recognize the scam while being manipulated.

• Truecaller can't help when scammers use spoofed numbers
• Bank fraud alerts catch transfers AFTER they happen
• The 1930 helpline requires victims to self-report (but 73% don't realize it's a scam until days later)
• Awareness campaigns assume people remember advice when they're under psychological duress

GuardianShield solves a different problem: detection, not prevention education.

We don't try to teach elderly people to spot scams under pressure. We empower their adult children to spot the behavioral patterns—long unknown calls + banking activity—and intervene before savings are transferred.

Because the 67-year-old woman in Hyderabad who lost ₹55 million over 17 days? She had family. They would have helped. They just didn't know she needed help.

GuardianShield exists to make sure someone knows in time to answer it.

What I learned:

• Technical feasibility matters more than features: Early designs had 24/7 call recording, but privacy concerns led us to duration-only tracking
• Banking API access is complex: Discovered RBI's Account Aggregator framework as the solution
• Cultural sensitivity is critical: Parents need autonomy, not surveillance— mutual consent and control features became central to design

The challenge: Building protection without violating dignity. Creating alerts that are actionable but not overwhelming. Integrating banking and telecom data that existing systems treat separately.

Built With

• 16-week
• 2.0
• a
• cloud-functions)-ml:-python-with-scikit-learn-for-pattern-detection-banking:-rbi-account-aggregator-framework-apis:-truecaller-api
• concept
• detailed
• implementation
• is
• oauth
• proposed-implementation-stack-(concept-stage):-android:-kotlin-with-accessibility-service-api-backend:-firebase-(firestore
• submission
• this
• whatsapp-business-api-security:-end-to-end-encryption
• with