Go Attendance — Smart Mobile Attendance for Schools

Go Attendance is an enterprise-grade mobile attendance management application built for educational institutions (teachers and school administrators). It modernizes traditional manual paper sheets and fixed fingerprint terminals into a secure, mobile-first workflow combining high-accuracy GPS geofencing, on-device face liveness verification, and server-authoritative time-window validation to prevent fraudulent attendance.

Schools and educational institutions struggled with outdated biometric terminals that created morning physical queues and lacked flexibility. Furthermore, conventional mobile check-in apps were susceptible to GPS spoofing, device clock tampering, and proxy check-ins. A resilient, tamper-proof system was required to guarantee authentic physical presence while keeping the check-in process under 20 seconds.

As the developer, I architected the cross-platform application using Clean Architecture (Feature-First) with Flutter and Riverpod. I implemented the complete Supabase backend infrastructure (PostgreSQL schemas, Row-Level Security policies, database triggers, and Deno Edge Functions) and integrated native device capabilities including Google ML Kit Face Detection, Geolocator, and Hive offline caching.

Client-side device clocks and GPS coordinates can be easily manipulated by users using mock location tools or manual time adjustments to bypass late thresholds.

Solusi: Designed a server-authoritative trust architecture. All attendance statuses (present, late, working_duration_min) and geofence validations are computed strictly within PostgreSQL triggers (derive_attendance_status()) using server time (now()) in the school's timezone. Client-submitted coordinates are re-evaluated server-side with Haversine distance formulas, ensuring zero reliance on untrusted client state.

Validating physical presence through selfies often causes slow check-in flows if image processing relies purely on external cloud APIs.

Solusi: Integrated Google ML Kit on-device face detection with a strict liveness validation pipeline (single face constraint, eye-open probability, and smiling ratio threshold). This ensures real-time sub-second verification directly on the device before compressing and uploading proof photos to Supabase Storage.

Teachers frequently forget to check out at the end of the day, leading to inaccurate duration logs and open-ended attendance records.

Solusi: Developed an automated background worker using Supabase Edge Functions (Deno) triggered via pg_cron. The scheduler dispatches push notification reminders between 20:00 and 04:00, and automatically resolves any unclosed daily records to auto_checkout status at 04:00 next day.

• Achieved a seamless check-in completion flow in under 15 seconds from cold app launch.
• Enforced strict 3-factor validation (GPS + Face Liveness + Server Time), eliminating proxy check-ins and mock location exploits.
• Delivered real-time administrative dashboards with sub-5-second sync using Supabase Realtime and Row Level Security (RLS).
• Provided 100% auditable attendance records backed by timestamped coordinates and facial evidence.