Scam Shield

A real-time voice deepfake detection progressive web app, included as evidence of practical AI safety product delivery rather than a healthcare deployment.

Problem

People need a usable signal when a live voice interaction may be synthetic or manipulated.

Users

People evaluating suspicious voice interactions.

Context

An AI safety prototype. It is not presented as a healthcare product or a perfect fraud-detection system.

Role

Designed and built the real-time voice deepfake detection PWA.

Requirements

• Analyse voice signals in real time
• Present the experience as an installable progressive web app

Approach

Packaged a real-time detection workflow into a focused web experience designed to surface a useful signal during a voice interaction.

Workflow

Voice input is analysed by the detection workflow and a result is presented to the user in the PWA.

Technology

Progressive web app, Real-time voice deepfake detection

Safety and risk considerations

Detection results should support human judgement, not be treated as certainty. The case study makes no accuracy or production-readiness claim.

Outcome

Produced a working PWA prototype for real-time voice deepfake detection.

Metrics

Not measured

Lessons

Safety-focused AI products need clear boundaries and a user experience that communicates uncertainty instead of hiding it.

Next steps

Define the intended operating conditions and evaluate detection quality against a representative test set.