Training face recognition model, image registration, face recognition method and device

Abstract

The embodiments of this specification provide a method and device for training a face recognition model, image registration, and face recognition to prevent privacy leakage. The method includes: the terminal device inputs the face training image into the first sub-model to obtain initial hidden layer features; uses the first public key to perform homomorphic encryption on the initial hidden layer features to obtain encrypted hidden layer features; sends the encrypted hidden layer features and The first public key; the server obtains the first encrypted output feature based on the encrypted hidden layer feature and the second sub-model; receives the first encrypted output feature from the server; uses the first private key to decrypt the first encrypted output feature to obtain the first decrypted Output feature; send the first decryption output feature to the server, and the server updates the second sub-model according to the first decryption output feature and the label of the face training image, and determines the gradient of the backpropagation to the first sub-model; receives the gradient from the server, The first sub-model is updated according to the gradient. It can improve the security of face recognition.

Description

technical field [0001] One or more embodiments of this specification relate to the field of computers, and in particular to methods and devices for training face recognition models, image registration, and face recognition. Background technique [0002] With the development of artificial intelligence and computer vision technology, face recognition technology, as a fast and convenient new authentication method, has been widely used in various fields, such as face payment in financial scenarios and face authentication in civil aviation scenarios , facial recognition access control in office scenarios, facial recognition authentication in hotel accommodation scenarios, etc. Compared with the traditional identity authentication methods based on swiping cards or passwords, face recognition has the advantages of non-contact, fast and effective, and difficult to forge, so it has gradually become the mainstream method of identity authentication. [0003] Face recognition technolog...

AI Technical Summary

Problems solved by technology

The current face recognition method still has a high risk of being attacked and leads to data leakage, and the security is low

Images