Whisper Shield: Offline-First AI Personal Safety Platform
An offline-first AI personal safety platform triggered by secret voice keywords, gestures, or SOS buttons to capture encrypted audio and broadcast GPS alerts without internet.
Stall Participants & Presenters
Overview
Over 80% of standard emergency SOS applications fail when the victim’s phone is locked, cellular internet coverage is lost, or when an attacker prevents direct access to the handset screen.
Whisper Shield is an offline-first, discreet personal safety ecosystem engineered to function reliably in adverse, high-stress environments. Developed for campus security, enterprise personnel protection, and municipal safety initiatives, the application guarantees emergency activation without cellular internet connectivity, contributing directly to SDG 5 (Gender Equality), SDG 11 (Sustainable Cities and Communities), and SDG 16 (Peace, Justice, and Strong Institutions).
Multi-Modal Trigger & Offline Defense Mechanisms
- Secret Acoustic Keyword Wakeup: Low-power acoustic inference scans audio every 30 seconds for the secret keyword phrase “Whisper Shield” across multiple regional accents and languages, bypassing device lock states.
- Physical Shake Gesture Detection: Continuous accelerometer telemetry detects calibrated dual-shake motions to trigger discreet emergency protocols while keeping the phone concealed in a bag or pocket.
- Manual Hardware & UI Override: Dedicated SOS activation button featuring a 5-second cancel buffer to prevent false alarms.
- Hybrid Vault Evidence Recording: Captures high-accuracy GPS coordinates, timestamps, and encrypted ambient audio evidence stored securely in a tamper-resistant local vault.
- Zero-Network Alert Transmission: Dispatches encrypted SMS emergency telemetry and offline Google Maps location vectors to designated guardians and emergency responders without requiring mobile data.
Architecture & Technology Stack
- Frontend Mobile App: Flutter and Dart cross-platform mobile client featuring native background services and low-power hardware sensor listeners.
- Backend Infrastructure: Node.js with Socket.io for real-time dispatch synchronizations when network connectivity is re-established.
- Audio Processing & AI: On-device acoustic keyword spotting with roadmap models for vocal distress emotion detection and scream anomaly classification.
- Security & Storage: AES-256 encrypted local data vault protecting audio recordings and telemetric evidence against unauthorized extraction.
- Integration Protocols: Telephony SMS API relays and WebSockets for low-latency emergency dispatch synchronization.

