# US 2021/0279441 A1 · Method and apparatus with anti-spoofing

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0279441 A1
- Document ID: US20210279441A1
- Status: Published application
- Kind: A1
- Application number: 17/172,429
- Filed: 2021-02-10
- Published: 2021-09-09
- Earliest claimed: 2020-03-05
- Assignees: Samsung Electronics Co., Ltd.
- Applicants: Samsung Electronics Co., Ltd.
- Inventors: Sungun PARK; Kyuhong KIM
- Web page: https://patentgrep.com/patent/US20210279441A1

## CPC

- G06V 40/1388: {using image processing}
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F 18/22: Matching criteria, e.g. proximity measures
- G06F 18/2413: based on distances to training or reference patterns
- G06F 21/32: using biometric data, e.g. fingerprints, iris scans or voiceprints
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06V 10/763: {Non-hierarchical techniques, e.g. based on statistics of modelling distributions}
- G06V 10/764: using classification, e.g. of video objects
- G06V 40/1347: {Preprocessing; Feature extraction}
- G06V 40/1365: {Matching; Classification}
- G06V 40/1382: {Detecting the live character of the finger, i.e. distinguishing from a fake or cadaver finger}
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A processor-implemented anti-spoofing method includes: extracting an input embedding vector from input biometric information; obtaining a fake embedding vector of a predetermined fake model based on fake biometric information; obtaining either one or both of a real embedding vector of a predetermined real model and an enrolled embedding vector of an enrollment model the enrollment model being generated based on biometric information of an enrolled user; determining a confidence value of the input embedding vector based on the fake embedding vector and either one or both of the real embedding vector and the enrolled embedding vector; and determining whether the input biometric information is forged, based on the confidence value.

## Claims

### Claim 1

A processor-implemented anti-spoofing method comprising:
extracting an input embedding vector from input biometric information;
obtaining a fake embedding vector of a predetermined fake model based on fake biometric information;
obtaining either one or both of a real embedding vector of a predetermined real model and an enrolled embedding vector of an enrollment model the enrollment model being generated based on biometric information of an enrolled user;
determining a confidence value of the input embedding vector based on the fake embedding vector and either one or both of the real embedding vector and the enrolled embedding vector; and
determining whether the input biometric information is forged, based on the confidence value.

### Claim 2 (depends on 1)

The method of claim 1, wherein the confidence value is determined based on a posterior probability between the input embedding vector and any one or any combination of the real embedding vector, the enrolled embedding vector, and the fake embedding vector.

### Claim 3 (depends on 2)

The method of claim 2, wherein the posterior probability is calculated based on a distance or a similarity between the input embedding vector and the any one or any combination of the real embedding vector, the enrolled embedding vector, and the fake embedding vector.

### Claim 4 (depends on 1)

The method of claim 1, wherein the determining of the confidence value comprises:
determining a value of a first category corresponding to a case in which the input biometric information is not forged, based on a first distance between the real embedding vector and the input embedding vector and a third distance between the enrolled embedding vector and the input embedding vector;
determining a value of a second category corresponding to a case in which the input biometric information is forged, based on a second distance between the fake embedding vector and the input embedding vector; and
determining the confidence value based on the value of the first category and the value of the second category.

### Claim 5 (depends on 4)

The method of claim 4, wherein the determining of the value of the first category comprises performing a weighted summation of the third distance and a lower value of the first distance and the third distance.

### Claim 6 (depends on 4)

The method of claim 4, wherein the determining of the confidence value based on the value of the first category and the value of the second category comprises applying the value of the first category and the value of the second category to a softmax function and determining the confidence value.

### Claim 7 (depends on 1)

The method of claim 1, wherein the determining of the confidence value comprises:
estimating either one or both of a first posterior probability between the input embedding vector and the real embedding vector and a third posterior probability between the input embedding vector and the enrolled embedding vector;
estimating a second posterior probability between the input embedding vector and the fake embedding vector; and
determining the confidence value based on the second posterior probability and the either one or both of the first posterior probability and the third posterior probability.

### Claim 8 (depends on 7)

The anti-spoofing method of claim 7, wherein:
the determining of the confidence value comprises calculating one or more probability ratios based on the second posterior probability and either one or both of the first posterior probability and the third posterior probability; and
the determining of whether the input biometric information is forged comprises determining whether the input biometric information is forged based on the one or more probability ratios.

### Claim 9 (depends on 8)

The method of claim 8, wherein
the calculating of the one or more probability ratios comprises:
calculating a first probability ratio between the first posterior probability and the second posterior probability; and
calculating a second probability ratio between the second posterior probability and the third posterior probability, and
the determining of the confidence value comprises performing a weighted summation of the first probability ratio and the second probability ratio.

### Claim 10 (depends on 1)

The method of claim 1, further comprising:
determining one or more embedding vectors closest to the input embedding vector among a plurality of embedding vectors included in any one or any combination of the real model, the fake model and the enrollment model, to be a representative embedding vector.

### Claim 11 (depends on 1)

The method of claim 1, wherein the fake embedding vector is an average value of a predetermined number of fake embedding vectors of the fake model that are closest to the input embedding vector.

### Claim 12 (depends on 1)

The method of claim 1, wherein
the extracting of the input embedding vector comprises extracting the input embedding vector from the input biometric information using a neural network that is trained to extract a feature for detecting whether biometric information is forged, and
the neural network is trained based on the real biometric information and the fake biometric information.

### Claim 13 (depends on 1)

The method of claim 1, wherein
the fake model corresponds to a set of fake embedding vectors extracted from the fake biometric information that is forged, and
the real model corresponds to a set of real embedding vectors extracted from unspecified real biometric information that is not forged.

### Claim 14 (depends on 1)

A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the method of claim 1.

### Claim 15

An anti-spoofing apparatus comprising:
a processor configured to:
extract an input embedding vector from input biometric information;
obtain a fake embedding vector of a predetermined fake model based on fake biometric information;
obtain either one or both of a real embedding vector of a real model and an enrolled embedding vector of an enrollment model the enrollment model being generated based on biometric information of an enrolled user;
determine a confidence value of the input embedding vector based on the fake embedding vector and either one or both of the real embedding vector and the enrolled embedding vector; and
determine whether the input biometric information is forged, based on the confidence value.

### Claim 16 (depends on 15)

The apparatus of claim 15, wherein the confidence value is determined based on a posterior probability between the input embedding vector and any one or any combination of the real embedding vector, the enrolled embedding vector, and the fake embedding vector

### Claim 17 (depends on 15)

The apparatus of claim 15, wherein, for the determining of the confidence value, the processor is configured to:
determine a value of a first category corresponding to a case in which the input biometric information is not forged, based on a first distance between the real embedding vector and the input embedding vector and a third distance between the enrolled embedding vector and the input embedding vector;
determine a value of a second category corresponding to a case in which the input biometric information is forged, based on a second distance between the fake embedding vector and the input embedding vector; and
determine the confidence value based on the value of the first category and the value of the second category.

### Claim 18 (depends on 15)

The apparatus of claim 15, wherein, for the determining of the confidence value, the processor is configured to:
estimate either one or both of a first posterior probability between the input embedding vector and the real embedding vector and a third posterior probability between the input embedding vector and the enrolled embedding vector;
estimate a second posterior probability between the input embedding vector and the fake embedding vector; and
determine the confidence value based on the second posterior probability and the either one or both of the first posterior probability and the third posterior probability.

### Claim 19 (depends on 18)

The apparatus of claim 18, wherein, for the determining of the confidence value, the processor is configured to:
calculate a first probability ratio between the first posterior probability and the second posterior probability;
calculate a second probability ratio between the second posterior probability and the third posterior probability; and
determine the confidence value based on the first probability ratio and the second probability ratio.

### Claim 20 (depends on 19)

The apparatus of claim 19, wherein, for the determining of the confidence value based on the first probability ratio and the second probability ratio, the processor is configured to perform a weighted summation of the first probability ratio and the second probability ratio.

### Claim 21 (depends on 15)

The anti-spoofing apparatus of claim 15, wherein the processor is configured to determine one or more embedding vectors closest to the input embedding vector among a plurality of embedding vectors included in any one or any combination of the real model, the fake model and the enrollment model, to be a representative embedding vector.

### Claim 22 (depends on 15)

The apparatus of claim 15, wherein
for the extracting of the input embedding vector, the processor is configured to extract the input embedding vector from the input biometric information using a neural network that is trained to extract a feature for detecting whether biometric information is forged, and
the neural network is trained based on the real biometric information and the fake biometric information.

### Claim 23 (depends on 15)

The apparatus of claim 15, comprising a sensor configured to sense the biometric information.

### Claim 24 (depends on 23)

The apparatus of claim 23, wherein the sensor is a fingerprint sensor and the biometric information corresponds to a fingerprint.

### Claim 25

A processor-implemented anti-spoofing method comprising:
extracting an input embedding vector from input biometric information;
determining a first, a second, and a third posterior probability respectively between the input embedding vector and a real embedding vector of a predetermined real model, a fake embedding vector of a predetermined fake model, and an enrolled embedding vector of an enrollment model of a user;
determining a confidence value of the input embedding vector based on a first probability ratio between the first posterior probability and the second posterior probability, and on a second probability ratio between the second posterior probability and the third posterior probability; and
determining whether the input biometric information is forged, based on the confidence value.
