Table 2 Computational limitation of traditional model comparison.
From: Secure facial biometric authentication in smart cities using multimodal methodology
Sl. No | Model | Encryption | Disadvantages | Proposed advantages |
|---|---|---|---|---|
1 | Standalone CNN | RSA | Due to the vanishing gradient problem, more computational effort is required to provide privacy using the RSA encryption technique. | The fused model performs well in handling gradient problems and requires low computational effort to process privacy using ElGamal encryption. |
2 | ResNet-50 | RSA | Processing the output takes longer time because it utilizes RSA with modular exponentiation, which requires larger key sizes. | The fused model utilizes CNN to reduce computational complexity and speed up the process, while ElGamal encryption offers a higher level of security than RSA. |
3 | VGG-16 | RSA | Compared to CNN and ResNet-50, a larger number of parameters are used which increases the computational complexity. | ResNet-50 uses minimum parameters when fused with CNN and has achieved better performance and enhanced privacy preservation using ElGamal encryption. |
