ARDIANSYAH, MUHAMMAD FARHAN (2026) IMPLEMENTASI PENGENALAN WAJAH MENGGUNAKAN FACENET DENGAN LIVENESS DETECTION BERBASIS MINIFASNETV2-SE UNTUK MENINGKATKAN KEAMANAN TERHADAP SERANGAN SPOOFING. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
COVER.pdf Download (790kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
BAB I.pdf Restricted to Registered users only Download (71kB) |
|
|
Text (BAB II)
BAB II.pdf Restricted to Registered users only Download (356kB) |
|
|
Text (BAB III)
BAB III.pdf Restricted to Registered users only Download (204kB) |
|
|
Text (BAB IV)
BAB IV.pdf Restricted to Registered users only Download (962kB) |
|
|
Text (BAB V)
BAB V.pdf Restricted to Registered users only Download (57kB) |
|
|
Text (DAFTAR PUSTAKA)
DAFTAR PUSTAKA.pdf Restricted to Registered users only Download (119kB) |
|
|
Text (LAMPIRAN)
LAMPIRAN.pdf Restricted to Registered users only Download (279kB) |
Abstract
Basic facial recognition systems like FaceNet are highly vulnerable to identity manipulation (spoofing) attacks using printed photos or digital screens. This study aims to enhance system security by integrating a MiniFASNetV2-SE based Liveness Detection model to verify facial authenticity prior to the identity extraction process. Test results revealed that the standalone FaceNet architecture had a fatal False Acceptance Rate for Spoofing (FAR Spoof) of 96.7%. However, by integrating MiniFASNetV2-SE, the FAR Spoof was drastically reduced to 8.3%, successfully blocking 100% of printed photo attacks. Evaluation of the Liveness Detection module yielded a total Attack Presentation Classification Error Rate (APCER) of 10.00%, a Bona Fide Presentation Classification Error Rate (BPCER) of 23.33%, and an Average Classification Error Rate (ACER) of 16.67%. In conclusion, this integration is proven effective in thwarting 2D spoofing attacks and significantly improves biometric authentication security. Key words : Face Recognition, FaceNet, Liveness Detection, MiniFASNetV2-SE, Spoofing, APCER, BPCER, FAR Spoof. Sistem pengenalan wajah dasar seperti FaceNet sangat rentan terhadap serangan manipulasi identitas (spoofing) menggunakan foto cetak maupun layar digital. Penelitian ini bertujuan meningkatkan keamanan sistem dengan mengintegrasikan model Liveness Detection berbasis MiniFASNetV2-SE untuk menyaring keaslian wajah sebelum tahap ekstraksi identitas. Hasil pengujian menunjukkan bahwa arsitektur FaceNet tunggal memiliki tingkat kebobolan (FAR Spoof) yang fatal sebesar 96,7%. Namun, setelah integrasi MiniFASNetV2-SE, FAR Spoof turun drastis menjadi 8,3%, dengan keberhasilan 100% memblokir serangan foto cetak. Evaluasi pada modul Liveness Detection mencatatkan Attack Presentation Classification Error Rate (APCER) total 10,00%, Bona Fide Presentation Classification Error Rate (BPCER) 23,33%, dan Average Classification Error Rate (ACER) 16,67%. Kesimpulannya, integrasi ini terbukti efektif menggagalkan serangan spoofing 2D dan secara signifikan meningkatkan keamanan otentikasi biometrik. Kata kunci : Pengenalan Wajah, FaceNet, Liveness Detection, MiniFASNetV2-SE, Spoofing, APCER, BPCER, FAR Spoof..
Actions (login required)
 |
View Item |
