Artificial intelligence-based identity verification method and apparatus, computer device, and storage medium

Patent metadata

Field	Value
Publication number	US-12424024-B2
Application number	US-202217976305-A
Country	US
Kind code	B2
Filing date	Oct 28, 2022
Priority date	Oct 16, 2020
Publication date	Sep 23, 2025
Grant date	Sep 23, 2025

Abstract

This application relates to an artificial intelligence-based identity verification method and apparatus, a computer device, and a storage medium. The method includes: obtaining a to-be-verified face image inputted into a verification device; performing artificial intelligence-based device tracing processing on the face image, to obtain a photographing device identifier corresponding to a photographing device of the face image; obtaining a verification device identifier corresponding to the verification device; performing feature comparison on the photographing device identifier and the verification device identifier to obtain a comparison result; and performing identity verification based on the face image in a case that the comparison result indicates that the photographing device and the verification device are a same device.

First claim

What is claimed is:

1. An artificial intelligence-based identity verification method, performed by a verification device, the method comprising:
obtaining a to-be-verified face image inputted into the verification device;
performing at least one convolution operation on the face image, to obtain an image convolution feature of the face image;
performing non-linear mapping on the image convolution feature, to obtain a first pattern noise feature;
performing artificial intelligence-based device tracing processing on the face image, to obtain a photographing device identifier corresponding to a photographing device of the face image based on the first pattern noise feature extracted from the face image;
obtaining a verification device identifier corresponding to the verification device based on a second pattern noise feature corresponding to the verification device;
determining a similarity between the first pattern noise feature and the second pattern noise feature;
performing feature comparison on the photographing device identifier and the verification device identifier to obtain a comparison result according to the similarity; and
performing identity verification based on the face image in response to the comparison result indicating that the photographing device and the verification device are a same device.

2. The method according to claim 1 , wherein the obtaining the comparison result of the photographing device identifier and the verification device identifier according to the similarity comprises:
obtaining a comparison result indicating that the photographing device and the verification device are a same device in response to the similarity being greater than a similarity threshold; and
obtaining a comparison result indicating that the photographing device and the verification device are different devices in response to the similarity being less than or equal to the similarity threshold.

3. The method according to claim 1 , wherein the performing non-linear mapping on the image convolution feature, to obtain the first pattern noise feature comprises:
performing pooling processing on the image convolution feature, to obtain an output result of the pooling; and
performing non-linear mapping on the output result of the pooling through an excitation function, to obtain the first pattern noise feature.

4. The method according to claim 1 , wherein the determining the photographing device identifier corresponding to the photographing device of the face image based on the first pattern noise feature comprises:
mapping the first pattern noise feature into a character string according to a device identifier format, to obtain the photographing device identifier corresponding to the photographing device of the face image; and
the performing feature comparison on the photographing device identifier and the verification device identifier to obtain a comparison result comprises:
performing character comparison on the photographing device identifier and the verification device identifier to obtain the comparison result.

5. The method according to claim 1 , wherein the device tracing processing is implemented based on a device tracing network model, the device tracing network model is generated through model training operations, and the model training operations comprise:
obtaining a training image carrying a device identifier label;
performing at least one convolution operation on the training image through a to-be-trained device tracing network model, to obtain a training convolution feature of the training image;
performing non-linear mapping on the training convolution feature through the device tracing network model, to obtain a training pattern noise feature;
determining a prediction device identifier corresponding to a photographing device of the training image through the device tracing network model according to the training pattern noise feature; and
continuing to perform training after adjusting a parameter of the device tracing network model according to the device identifier label and the prediction device identifier, until a trained device tracing network model is obtained after the training is ended.

6. The method according to claim 5 , wherein the continuing to perform training after adjusting a parameter of the device tracing network model according to the device identifier label and the prediction device identifier comprises:
obtaining a distance loss between the device identifier label and the prediction device identifier;
determining a gradient parameter of the distance loss; and
adjusting the parameter of the device tracing network model according to the gradient parameter, and continuing to perform training according to the device tracing network model whose parameter is adjusted.

7. The method according to claim 1 , wherein the obtaining a verification device identifier corresponding to the verification device comprises:
obtaining a reference image photographed by the verification device; and
performing artificial intelligence-based device tracing processing on the reference image, to obtain the verification device identifier corresponding to the verification device.

8. The method according to claim 1 , wherein the performing identity verification based on the face image comprises:
obtaining a verification standard image;
performing face comparison on the face image and the verification standard image, to obtain a face comparison result; and
obtaining an identity verification result of the face image according to the face comparison result.

9. The method according to claim 1 , wherein the method further comprises:
in response to the comparison result indicating that the photographing device and the verification device are different devices, instructing the verification device to display a verification failure prompt message, and instructing the verification device to display prompt information describing that the face image is an illegal image.

10. An artificial intelligence-based identity verification apparatus, the apparatus comprising:
a memory storing a plurality of instructions; and
a processor configured to execute the plurality of instructions, and upon execution of the plurality of instructions, is configured to:
obtain a to-be-verified face image inputted into a verification device;
perform artificial intelligence-based device tracing processing on the face image, to obtain a photographing device identifier corresponding to a photographing device of the face image, wherein the device tracing processing is implemented based on a device tracing network model, the device tracing network model is generated through a model training operations, and the model training operations comprise:
an operation to obtain a training image carrying a device identifier label;
an operation to perform at least one convolution operation on the training image through a to-be-trained device tracing network model, to obtain a training convolution feature of the training image;
an operation to perform non-linear mapping on the training convolution feature through the device tracing network model, to obtain a training pattern noise feature;
an operation to determine a prediction device identifier corresponding to a photographing device of the training image through the device tracing network model according to the training pattern noise feature; and
an operation to continue to perform training after adjusting a parameter of the device tracing network model according to the device identifier label and the prediction device identifier, until a trained device tracing network model is obtained after the training is ended;

obtain a verification device identifier correspondin

Assignees

• Tencent Tech Shenzhen Co Ltd

Inventors

• Luo Zhao
• Bai Kun

Classifications

• G06V10/82

  using neural networks · CPC title

• G06F21/44

  Program or device authentication · CPC title

• H04L63/0861

  using biometrical features, e.g. fingerprint, retina-scan (cryptographic mechanisms or cryptographic arrangements for entity authentication using biological data H04L9/3231) · CPC title

• G06V40/172 Primary

  Classification, e.g. identification · CPC title

• G06V40/166

  using acquisition arrangements · CPC title