Stateless Geometric Watermarking for Replay-Resistant QR-Based Attendance Verification

Abstract

QR codes are used in many educational institutions for attendance tracking. This is due to their low cost, rapid deployment, and compatibility with mobile devices. However, a drawback of using QR codes is their ease of resending. A student can take a picture of the code and send it to a friend, who can then use it without physically attending. This leads to problems with the accuracy of the attendance system. This study proposes a replay-resistant QR attendance framework that verifies both the decoded payload and the visual provenance of the submitted QR image. The method embeds a lightweight directional geometric watermark into perceptually stable QR modules using controlled one-pixel perturbations. A deterministic redundant mapping strategy derives embedding coordinates from the session context, enabling stateless server-side verification without storing image-specific templates or reference maps. The framework was implemented as a .NET microservice integrated with Flutter teacher and student applications and evaluated under authentic classroom capture, screenshot replay, WhatsApp forwarding, and Telegram forwarding scenarios using 100 heterogeneous smartphones and more than 1000 verification trials. Original captures achieved an average watermark matching rate of 0.94, while replay-oriented transmissions recorded lower rates due to degradation of embedded geometric signals resulting from re-compression, interpolation, scaling, and re-encoding. At an operational threshold of 0.75, the system achieved an accuracy of 96.30%, a false accept rate of 4.17%, and a false reject rate of 3.00%, with an area under the curve (AUC) of approximately 0.985. These results indicate that geometric watermarks can provide a practical and low-cost security layer for QR code-based attendance systems, without requiring specialized hardware, advanced forensic models, or reference databases for each image.