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

By training the encrypted model and face recognition model, the face image is converted into a noisy image style, which solves the problem of privacy protection and accuracy reduction during the use of face images, and achieves the accuracy of face recognition while protecting privacy while maintaining accuracy of face recognition sex.

CN113592696BActive Publication Date: 2024-11-12ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110927877.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-12
Publication Date
2024-11-12
Estimated Expiration
2041-08-12

AI Technical Summary

Technical Problem

The prior art is difficult to maintain the accuracy of execution of downstream tasks while protecting user privacy during the use of face images, especially when training of face recognition models, the accuracy of encrypted face images may drop significantly.

Method used

By training the encryption model and the face recognition model, the encrypted face image is converted into a noisy image style using a generative adversarial network, combined with the discriminant model, the target of minimizing the total prediction loss, the encrypted model is trained to ensure that the encrypted face image is closer to the noisy image. , protect privacy while maintaining the accuracy of identification results.

Benefits of technology

It realizes the accuracy of the recognition results of the face recognition model while protecting user privacy, which is comparable to the recognition results obtained from the training of the original face image, ensuring that the facial features in the encrypted face image are not leaked and do not affect subsequent tasks. accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113592696B_ABST
    Figure CN113592696B_ABST
Patent Text Reader

Abstract

The embodiments of this specification provide 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. The encryption model training method includes: using the encryption model to be trained to encrypt the original face image to obtain an encrypted face image; inputting the encrypted face image into the first face recognition model to be trained to obtain a first face prediction result; using the first face prediction result and the face annotation result corresponding to the original face image to determine the first prediction loss; inputting the encrypted face image into the discriminant model, and obtaining the first prediction probability that the encrypted face image is a noise image through the discriminant model; determining the total prediction loss based on a positive correlation with the first prediction loss and a negative correlation with the first prediction probability; and training the encryption model and the first face recognition model with the goal of minimizing the total prediction loss.
Need to check novelty before this filing date? Find Prior Art