<![CDATA[The surge in vehicle theft in Indonesia has exposed the weaknesses of conventional security systems such as mechanical keys and immobilizers. This study aims to develop an IoT-based vehicle security system integrating fingerprint authentication, facial recognition, and GPS tracking, and evaluates its performance quantitatively. The system was tested on a Toyota Avanza with six users. Facial recognition used a MobileNetV2 CNN model trained with 1,200 local images across four classes (registered, unregistered, masked, and sunglasses) using a learning rate of 0.001, batch size of 32, and 50 epochs. Fingerprint authentication employed minutiae extraction with Euclidean distance matching. The system successfully implemented two-factor authentication. Facial recognition achieved an accuracy of 94.2%, with a False Acceptance Rate (FAR) of 2.1% and a False Rejection Rate (FRR) of 3.7%. Fingerprint authentication reached 91.5% accuracy, with FAR of 4.3% and FRR of 4.2% under dry finger conditions, while FRR increased to 18.5% for scratched fingers. The system detected unregistered users and triggered engine shutdown while sending photos and GPS coordinates through Telegram. GPS tracking achieved 99.3% positional accuracy. The results demonstrate the feasibility of multimodal IoT-based vehicle security, although performance remains sensitive to lighting conditions, face coverings, and finger surface conditions.]]>
Copyrights © 2026
