Feature Interaction-Based Face De-Morphing Factor Prediction for Restoring Accomplice's Facial Image.

Face morphing attacks disrupt the essential correlation between a face image and its identity information, posing a significant challenge to face recognition systems. Despite advancements in face morphing attack detection methods, these techniques cannot reconstruct the face images of accomplices. Existing deep learning-based face de-morphing techniques have mainly focused on identity disentanglement, overlooking the morphing factors inherent in the morphed images. This paper introduces a novel face de-morphing method to restore the identity information of accomplices by predicting the corresponding de-morphing factor. To obtain reasonable de-morphing factors, a channel-wise attention mechanism is employed to perform feature interaction, and the correlation between the morphed image and the real-time captured reference image is integrated to promote the prediction of the de-morphing factor. Furthermore, the identity information of the accomplice is restored by mapping the morphed and reference images into the StyleGAN latent space and performing inverse linear interpolation using the predicted de-morphing factor. Experimental results demonstrate the superiority of this method in restoring accomplice facial images, achieving improved restoration accuracy and image quality compared to existing techniques.

Security improvement for asymmetric cryptosystem based on spherical wave illumination.

We propose an improvement method for an asymmetric cryptosystem based on spherical wave illumination. Compared with the phase-truncated Fourier transform-based cryptosystem and the reported improving methods, the encryption process uses a spherical wave to illuminate the encryption system, rather than a uniform plane wave. As a result, the proposed method can avoid various types of the currently existing attacks and maintain the asymmetric characteristic of the cryptosystem. Moreover, due to only changing the illuminating mode, the proposed method can be easily implemented in optics compared with the reported improving methods. Simulation results are presented to demonstrate the feasibility and the security performance of the proposed method.