Security Architecture, Biometrics & Anti-Spoofing
In-depth technical breakdown of data encryption, on-device biometric vectors, and location spoofing boundaries.
1. On-Device Facial Vector Extraction
The single greatest hazard in biometric attendance is a database breach exposing employee photos. CloudAttend eliminates this vulnerability at the client architectural level:
↳ Raw Camera Frame Discarded from RAM Immediately
↳ Only 128 Floats Transmitted via TLS 1.3 to Database
Upon check-in, matching calculates the Euclidean distance between vectors: √(∑(Aᵢ - Bᵢ)²). If the distance falls below the 0.65 tolerance threshold, verification succeeds.
2. Facial Anti-Spoofing & Liveness Boundaries
We state plainly: the current codebase computes numerical vector similarity and does not currently include active 3D liveness detection (such as infrared depth analysis, blink detection, or head turn challenges). A printed photograph held before an unsupervised lens could produce a matching vector. CloudAttend mitigates this by requiring simultaneous workplace GPS geofencing to prevent remote proxy clock-ins.
3. Browser Geofencing & Mock Location Boundaries
HTML5 web browsers are sandboxed by Android and iOS for security. Consequently, web browsers cannot query operating system kernel flags to detect whether a third-party mock location provider (Fake GPS app) is running on the device.
Our mitigation: CloudAttend logs full coordinate precision. Static Fake GPS coordinates repeated to seven decimal places on consecutive days stand out immediately during supervisor audits.
4. Network Hygiene, Rate Limiting & Bot Defense
The codebase implements edge middleware that drops automated vulnerability probes, applies IP-based rate limiting against login brute forcing, and signs session tokens using Web Crypto JWT standards.