Edukasi Perlindungan Data Biometrik melalui Pengenalan YOLOv12 dan FastSAM bagi Mahasiswa

Abstract

Face recognition requires an understanding of application functions and biometric data risks. This community service program at STIE Hidayatullah Depok aimed to support students’ biometric literacy and identify further educational needs. The activity took place on October 11, 2026, from 08:00 to 13:00 Western Indonesian Time through presentations and discussion. Evaluation covered 19 pre-test and 17 post-test responses, each containing five questions worth 20 points, with 15 matched respondents. Analysis comprised score verification, descriptive statistics, and item-level response accuracy. Mean scores were 89.47 ± 18.10 and 77.65 ± 19.85, respectively, and were reported separately because the instruments assessed different content. Post-test accuracy reached 100.00% for unknown-label interpretation, 94.12% for FastSAM, 76.47% for embeddings, 64.71% for YOLOv12, and 52.94% for liveness. Four of the 12 responses scoring 80–100 still answered the liveness question incorrectly. These findings provide a basis for prioritizing content for students at the partner institution. Assessable outcomes are limited to item-level knowledge and identified learning needs. Knowledge improvement, behavioral change, and practical skills cannot yet be established. Proposed follow-up focuses on liveness and detection functions, supported by equivalent instruments for future evaluation.