Decoupling Identity from Ballot: A Tokenised Privacy Model for CPU-Based Biometric Voting

Abstract

Paper rolls and manual ID verification are still in use for many institutional elections, whose solution is easily overcome. SmartGov is an electronic voting system that closes the gap without any special hardware. It works on a normal CPU and requests the voter to provide three things: a match of a face, a live video stream, and, finally, a confirmation that they're the right individual. The system sends a one-time code to the voter's cell phone and requires a gesture. Face detection is implemented through InsightFace SCRFD, and ArcFace ResNet-50 implements recognition. Gesture Verification uses MediaPipe Face Mesh to run in the browser. There is no database outside the system; there is no biometric terminal or any GPU. Voters' confidence in the separation of their identity from the ballot is maintained until the ballot is cast, after which the token is destroyed. Among 100 participants, all of the spoofing and duplicate voting attempts were rejected, and the average time spent on authentication was 1.8 seconds. An institution without biometric infrastructure can implement SmartGov, a readily available security upgrade.