Encryption model training method and device, image encryption method and device and encrypted face image recognition method and device

Abstract

The embodiment of the invention provides an encryption model training method and device, an image encryption method and device and an encrypted face image recognition method and device. The encryption model training method comprises the steps: encrypting an original face image through a to-be-trained encryption model, and obtaining an encrypted face image; inputting the encrypted face image into a to-be-trained first face recognition model to obtain a first face prediction result; determining a first prediction loss by using the first face prediction result and a face labeling result corresponding to the original face image; inputting the original face image and the encrypted face image into a trained second face recognition model to obtain a first output result and a second output result corresponding to the original face image and the encrypted face image; determining a second prediction loss based on the first output result and the second output result; determining a total predicted loss based on a positive correlation with the first predicted loss and the second predicted loss; and training an encryption model and a first face recognition model by taking minimization of the total prediction loss as a target.

Description

technical field [0001] This specification relates to the technical field of image processing, and in particular to an encryption model training method and device for privacy protection, an image encryption method and device, and an encrypted face image recognition method and device. Background technique [0002] In recent years, data security and personal privacy protection have attracted more and more attention from the public. Due to the unpredictability of the security of the operating environment of third-party equipment, the security of data in the process of using face images, such as face recognition applications, is greatly threatened. How to protect the privacy of users in the whole process of using face images is an issue worthy of research and discussion. [0003] At present, the process of protecting the privacy of users in the whole process of using face images is generally: using the encryption algorithm in traditional digital image processing methods to encry...

AI Technical Summary

Problems solved by technology

[0004] However, in the above process, the encrypted face image can protect the user's personal privacy to a certain extent, but it cannot control the impact of the image feature loss of the encrypted face image on downstream tasks, such as the encrypted face image. Accuracy may drop significantly when images are used to train face recognition models

Images