Biological characteristic encryption and identification method for privacy protection
By encrypting and restoring biometric images using denoising diffusion implicit model, and using original and restored images in the recognition model, the contradiction between the protection of privacy and the recognition accuracy of biometric encryption methods is solved, and efficient and accurate identity recognition and privacy protection are achieved.
CN120197161APending Publication Date: 2025-06-24SOUTH CHINA UNIV OF TECH
0 Cites 2 Cited by
Patent Information
- Application Number
- CN202510119285.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-24
- Publication Date
- 2025-06-24
AI Technical Summary
Technical Problem
Existing biometric encryption methods are difficult to achieve efficient and accurate identity recognition while protecting personal privacy.
Method used
The reverse noise addition and forward denoising processes of the denoising diffusion implicit model are used to encrypt and restore the biometric images, and the original image and the restored image are used for subsequent training of the recognition model.
Benefits of technology
It realizes that the identity identification can be efficiently and accurately after encryption, while effectively protecting personal privacy and preventing forgery and malicious attacks.
✦ Generated by Eureka AI based on patent content.
Abstract
The invention discloses a privacy protection biological feature encryption and identification method. The method comprises the following steps: pre-training a denoising diffusion implicit model; encrypting and recovering samples in the original biological characteristic image data set by using a reverse noise adding process and a forward denoising process of the trained denoising diffusion implicit model; inputting the original image and the recovered image of the data set into a to-be-trained recognition model, and carrying out joint training; in practical application, after a client obtains a biological feature image to be recognized, the image is encrypted by using a reverse noise adding process of a trained denoising diffusion implicit model; the encrypted image is transmitted to a server, and the server recovers the image by using a forward denoising process of a denoising diffusion implicit model; inputting the recovered image into a recognition model to obtain a feature vector; and after feature comparison, an identification result is transmitted back to the client. According to the method, the efficient and accurate recognition effect can be ensured while the personal privacy is protected.
Need to check novelty before this filing date? Find Prior Art