Tehnički vjesnik, Vol. 32 No. 5, 2025.
Izvorni znanstveni članak
https://doi.org/10.17559/TV-20240801001893
GAN-Based Facial Information Protection for IoT Using Transfer Probability Models
Ye Qin
; Vocational and Technical College, Inner Mongolia Agricultural University, 010018, China
Yaowu Kang
; Vocational and Technical College, Inner Mongolia Agricultural University, 010018, China
*
* Dopisni autor.
Sažetak
Facial information in the Internet of Things (IoT) faces great security challenges. This study proposes a novel facial information protection model combining Deep Convolutional Generative Adversarial Networks (DCGAN) with Transfer Probability Models (TPM). The model generates high-quality virtual face images while preserving key features of the original image. The results demonstrated that the model performed well in terms of encryption-decryption error (0.05), speed (632.5 Mbit/s for encryption and 583.5 Mbit/s for decryption), resource consumption (21.1%), and latency (11.5%) compared with existing methods. The model achieved more than 90% privacy protection for identity, facial expression, shape, and gesture, proving its effectiveness in facial information protection for IoT.
Ključne riječi
deep convolutional generative adversarial network; facial features; information; protection transition probability model
Hrčak ID:
335058
URI
Datum izdavanja:
30.8.2025.
Posjeta: 504 *