CN109753921A - A Face Feature Vector Privacy-Preserving Recognition Method - Google Patents
A Face Feature Vector Privacy-Preserving Recognition Method Download PDFInfo
- Publication number
- CN109753921A CN109753921A CN201811640575.4A CN201811640575A CN109753921A CN 109753921 A CN109753921 A CN 109753921A CN 201811640575 A CN201811640575 A CN 201811640575A CN 109753921 A CN109753921 A CN 109753921A
- Authority
- CN
- China
- Prior art keywords
- face
- uproar
- feature vector
- num
- secret protection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000013598 vector Substances 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000001815 facial effect Effects 0.000 claims description 25
- 238000012545 processing Methods 0.000 claims description 16
- 230000003044 adaptive effect Effects 0.000 claims description 10
- 238000012706 support-vector machine Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000003760 hair shine Effects 0.000 claims 1
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 238000013500 data storage Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Landscapes
- Collating Specific Patterns (AREA)
Abstract
本发明公开了一种人脸特征向量隐私保护识别方法，涉及人脸识别领域，包括如下步骤：S10，参考对象人脸图像隐私保护；S20，被识别对象人脸图像隐私保护；S30，对所述被识别对象进行识别。本发明降低了服务提供商数据存储上的压力和因为数据泄露而带来的道德与法律风险，避免了网络传输中不安全因素所造成的脸部隐私泄露风险，有效地保护了人脸识别的隐私。
The invention discloses a face feature vector privacy protection recognition method, which relates to the field of face recognition and includes the following steps: S10, the privacy protection of the face image of the reference object; S20, the privacy protection of the face image of the recognized object; S30, the The identified object is identified. The invention reduces the pressure on service provider data storage and the moral and legal risks caused by data leakage, avoids the risk of face privacy leakage caused by unsafe factors in network transmission, and effectively protects the face recognition privacy.
Description
Technical field
The present invention relates to field of face identification more particularly to a kind of face feature vector secret protection recognition methods.
Background technique
With the development of artificial intelligence deep learning, the promotion of raising and the capital market of processor speed, face is known
Application scenarios not in life are more and more.Recognition of face can fast and effeciently carry out personnel's verification, significantly avoid because
Material fakes, falsely uses identity etc. to inconvenience caused by personal and mechanism and loss.However recognition of face also brings privacy of user
The problem of.Currently, recognition of face service depends on the photo for uploading user and camera catcher's face picture all
It sends to and third party's enterprise platform of recognition of face service is provided, the processing and matching of picture are carried out by them.These platforms money
Matter is irregular, and the protection of privacy of user is also difficult to supervise.Therefore service user is uploaded to offer face in image
The privacy of itself face image is not just can guarantee after the enterprise platform of identification service.Meanwhile the malice of face image is passed
It broadcasts and is likely to bring the problems such as information security, internet-relevant violence, portrait infringement.
It, can be to avoid references object using the face feature vector secret protection recognition methods for introducing self-adapting random noise
Directly be uploaded to perhaps insecure third party's enterprise platform with identified subject face image, selection camera it is local into
The processing of row image.On the other hand, it joined random life in the face original feature vector that face characteristic calculates network output
At the adaptive noise of, Gaussian distributed, even so that same picture, every time by face characteristic calculating network after
The feature vector of output is still different.This behave, which is reduced, carries out inverse operation to feature vector to obtain raw facial
A possibility that image.Finally when service provider carries out recognition of face, storage is to add feature vector of making an uproar, and compares identification object
It also is to add feature vector of making an uproar.Service provider can not directly contact the face picture of service user, can not also obtain from currently
Take plus feature vector inverse operation of making an uproar obtains the face image of script, to ensure that the face image privacy of service user.
It finds by prior art documents, there is presently no use the above method to realize the face of secret protection
Identifying schemes.
Therefore, those skilled in the art is dedicated to developing a kind of face feature vector secret protection recognition methods.
Summary of the invention
In view of the above drawbacks of the prior art, the technical problem to be solved by the present invention is to how realize secret protection
Recognition of face.
To achieve the above object, the present invention provides a kind of face feature vector secret protection recognition methods.In the present invention,
The service provider for providing recognition of face will cannot directly acquire the face image of references object and identified object, and service mentions
It is not enough to carry out the face image that inverse operation obtains script for the information that quotient obtains.User, that is, references object, which uses, has face special
The camera for levying vector secret protection recognition methods shoots facial photo, locally carries out the place for face image in camera
Reason, processing include calculating the original feature vector that network obtains one group of characterization facial characteristics by face characteristic, are added by the vector
It is upper random generate, obtain a N-dimensional after the adaptive noise of Gaussian distributed plus the feature vector M that makes an uproarRNoiseFace.Then
It is waken up by far call interface and the service provider of recognition of face service is provided, service provider will add feature vector of making an uproar
MRNoiseFaceIt is put into wait in database and is matched identification.Similarly, if there are secret protection needs in identification process, known
Side does not need to dispose the camera with face feature vector secret protection recognition methods, and local detection simultaneously intercepts face image,
Calculate specified identified object adds the feature vector M that makes an uproarINoiseFaceAnd the references object M that will be stored with service providerRNoiseFace
By comparing determining device.Multilevel iudge device utilizes SVM (Support Vector Machine) or Euclidean distance threshold determination, most
The matching result of output face eventually.When parameter setting is reasonable, it still is able to keep the recognition of face precision of script using this method.
Specifically, a kind of face feature vector secret protection recognition methods includes the following steps:
S10, references object facial image secret protection；
S20 is identified the secret protection of object facial image；
S30 identifies the identified object.
Further, the references object facial image secret protection includes:
S101 shoots the references object face using the camera with references object facial image privacy protection function
Photo；
S102 calculates the original feature vector M of references object face described in network query function by face characteristicRFace；
S103, to the original feature vector M of the references object faceRFaceCarry out plus make an uproar processing, calculates the reference pair
Add the feature vector M that makes an uproar as faceRNoiseFace；
The references object face is added the feature vector M that makes an uproar by S104RNoiseFaceIt is saved in database.
Further, the step S104 includes:
S1040 utilizes far call interface call service end；
The references object face is added the feature vector M that makes an uproar by S1041RNoiseFaceIt is transferred to the server-side；
S1042, the server-side add the feature vector M that makes an uproar for the references object faceRNoiseFaceIt is stored in database
In and clustered.
Further, the identified object facial image secret protection includes:
S201 uses the camera shooting identified object with identified object facial image privacy protection function
Facial photo；
S202 calculates the original feature vector M of the identified subject faceIFace；
S203, to the original feature vector M of the identified subject faceIFaceCarry out plus make an uproar processing, calculates identified pair
Add the feature vector M that makes an uproar as faceINoiseFace。
Further, carrying out identification to the identified object includes:
S301, the identified subject face add the feature vector M that makes an uproarINoiseFaceWith the references object face plus make an uproar
Feature vector MRNoiseFaceIt is compared；
S302 exports the face matching result of the identified object.
Further, the step S103 adds device of making an uproar to carry out adding processing of making an uproar using adaptive Gauss, and the Gauss, which adds, to make an uproar
Device includes the noise coefficient matrix M of pre-set N*NRk, default N-dimensional variance vectors MRσ_default, shooting face number
The N-dimensional variance vectors M gone out according to statisticsRσ_statWith the MRkDiagonal line on N number of noise figure kRn；When the face number of shooting is small
When the lower limit Rsmall_num of references object face number, that is, when the face number < Rsmall_num shot, MRσ=0.8*
MRσ_default+0.2*MRσ_stat；When the face number of shooting is greater than the lower limit Rsmall_num of references object face number and claps
When the face number taken the photograph is less than the larger threshold value Rbig_num of references object face number, that is, face number > Rsmall_ for shooting
Num and when < Rbig_num, MRσ=0.5*MRσ_default+0.5*MRσ_stat；When the face number of shooting is greater than references object face
When the larger threshold value Rbig_num of number, that is, M when the face number > Rbig_num shotRσ=0.2*MRσ_default+0.8*
MRσ_stat；The M as the face number of shooting > > Rbig_num (the face number > 10*Rbig_num shot)Rσ=MRσ_stat。
Further, above-mentioned MRkDiagonal line on N number of noise figure kRn, in Gaussian Profile (MRk*MRσ) in can multiply as value
Get up, the specific value of N number of noise figure can be set to different value, respectively according to identification plus the demand made an uproar on diagonal line
(kR1, kR2..., kRn)。
Further, the noise vector M that the Gauss adds the device Gauss that makes an uproar that the every wheel of device of making an uproar is added to generate one group of N-dimensionalRNoise=Gauss
Distribution (0, MRk*MRσ), by the MRFaceValue and the M on often one-dimensionalRNoiseIt is added, generate plus make an uproar feature vector
MRNoiseFace。
Further, the step S203 includes that device of making an uproar is added to carry out adding processing of making an uproar, the Gauss using adaptive Gauss
Adding device of making an uproar includes the noise coefficient matrix M of pre-set N*NIk, default N-dimensional variance vectors MIσ_default, shooting people
The N-dimensional variance vectors M that face data statistics goes outIσ_statWith the MIkDiagonal line on N number of noise figure kIn；When the face number of shooting
When mesh is less than the lower limit Ismall_num of references object face number, that is, when the face number < Ismall_num shot, MIσ=
0.8*MIσ_default+0.2*MIσ_stat；When the face number of shooting is greater than the lower limit Ismall_num of references object face number simultaneously
And the face number of shooting is when being less than the larger threshold value Ibig_num of references object face number, that is, the face number that shoots >
Ismall_num and when < Ibig_num, MIσ=0.5*MIσ_default+0.5*MIσ_stat；When the face number of shooting is greater than reference
When the larger threshold value Ibig_num of subject face number, that is, M when the face number > Ibig_num shotIσ=0.2*MIσ_default+
0.8*MIσ_stat；The M as the face number of shooting > > Ibig_num (the face number > 10*Ibig_num shot)Iσ=
MIσ_stat。
Further, above-mentioned MIkDiagonal line on N number of noise figure kIn, in Gaussian Profile (MIk*MIσ) in can multiply as value
Get up, the specific value of N number of noise figure can be set to different value, respectively according to identification plus the demand made an uproar on diagonal line
(kI1, kI2..., kIn)。
Further, the noise vector M that the Gauss adds the device Gauss that makes an uproar that the every wheel of device of making an uproar is added to generate one group of N-dimensionalINoise=Gauss
Distribution (0, MIk*MIσ), by the MIFaceValue and the M on often one-dimensionalINoiseIt is added, generate plus make an uproar feature vector
MINoiseFace。
Further, the step S301 includes:
The identified subject face is added the feature vector M that makes an uproar by S3010INoiseFaceWith the references object face
Add the feature vector M that makes an uproarRNoiseFaceInput multilevel iudge device；
S3011, multilevel iudge device utilize SVM (Support Vector Machine) or Euclidean distance threshold value, judge institute
That states identified subject face adds the feature vector M that makes an uproarINoiseFaceWith the references object face respectively clustered in the database
Plus make an uproar feature vector MRNoiseFaceSimilarity degree.
The invention has the following beneficial effects:
(1) service provider for providing recognition of face service does not need that references object can not be directly acquired and is identified
The face image of object；
(2) adaptive Gauss is added in local pretreatment and adds device of making an uproar, (i.e. pond is from facial image in pooling
Extract a kind of mathematical method for being taken during face characteristic) on the basis of further reduced inverse behaviour is carried out to feature vector
Make a possibility that obtaining raw facial image；
(3) facial eigenvectors after making the service provider for providing recognition of face service only store plus make an uproar.Even if letting out
Database information is revealed, the face image of service user still can ensure privacy.Simultaneously as the management of camera, people
The reception preservation of face image is locally determined by user, reduces the pressure in service provider data storage and because data are let out
Reveal and bring morals and legal risk.
(4) whole processing that the present invention refers to are all without departing from the local machine where camera, such as video camera, mobile phone, a
People's computer etc., it is entirely avoided face's privacy leakage risk caused by insecurity factor in network transmission.
It is described further below with reference to technical effect of the attached drawing to design of the invention, specific structure and generation, with
It is fully understood from the purpose of the present invention, feature and effect.
Detailed description of the invention
Fig. 1 is the secret protection schematic diagram of the references object face image of a preferred embodiment of the invention；
Fig. 2 is the secret protection schematic diagram of the identified subject face image of a preferred embodiment of the invention；
Fig. 3 is the schematic diagram applied face feature vector secret protection recognition methods and carry out recognition of face.
Specific embodiment
Multiple preferred embodiments of the invention are introduced below with reference to Figure of description, keep its technology contents more clear and just
In understanding.The present invention can be emerged from by many various forms of embodiments, and protection scope of the present invention not only limits
The embodiment that Yu Wenzhong is mentioned.
In the accompanying drawings, the identical component of structure is indicated with same numbers label, everywhere the similar component of structure or function with
Like numeral label indicates.The size and thickness of each component shown in the drawings are to be arbitrarily shown, and there is no limit by the present invention
The size and thickness of each component.Apparent in order to make to illustrate, some places suitably exaggerate the thickness of component in attached drawing.
A preferred embodiment of the invention, comprising:
S10, references object facial image secret protection；
S20 is identified the secret protection of object facial image；
S30 identifies the identified object.
As shown in Figure 1, before references object uses recognition of face to service for the first time, first carry out the analysis of face image and
Data upload, that is, register.References object facial image secret protection includes:
S101 shoots the references object face using the camera with references object facial image privacy protection function
Photo；
S102 calculates the original feature vector M of references object face described in network query function by face characteristicRFace；
S103, to the original feature vector M of the references object faceRFaceCarry out plus make an uproar processing, calculates the reference pair
Add the feature vector M that makes an uproar as faceRNoiseFace；Device of making an uproar is added to carry out adding processing of making an uproar using adaptive Gauss, the Gauss, which adds, to make an uproar
Device includes the noise coefficient matrix M of pre-set N*NRk, default N-dimensional variance vectors MRσ_default, shooting face number
The N-dimensional variance vectors M gone out according to statisticsRσ_statWith the MRkDiagonal line on N number of noise figure kRn；When shooting face number <
When Rsmall_num, MRσ=0.8*MRσ_default+0.2*MRσ_stat；As the face number>Rsmall_num and<Rbig_ of shooting
When num, MRσ=0.5*MRσ_default+0.5*MRσ_stat；The M as the face of shooting number > Rbig_numRσ=0.2*MRσ_default
+0.8*MRσ_stat；The M as the face number of shooting > > Rbig_num (the face number > 10*Rbig_num shot)Rσ=
MRσ_stat；The noise vector M that the Gauss adds the device Gauss that makes an uproar that the every wheel of device of making an uproar is added to generate one group of N-dimensionalRNoise=Gaussian Profile (0, MRk*
MRσ), by the MRFaceValue and the M on often one-dimensionalRNoiseIt is added, generate plus make an uproar feature vector MRNoiseFace；
The references object face is added the feature vector M that makes an uproar by S104RNoiseFaceIt is saved in database, specifically includes:
S1040 utilizes far call interface call service end；
The references object face is added the feature vector M that makes an uproar by S1041RNoiseFaceIt is transferred to the server-side；
S1042, the server-side add the feature vector M that makes an uproar for the references object faceRNoiseFaceIt is stored in database
In and clustered.
As shown in Fig. 2, after service provider establishes characteristic vector space database, such as enter brush face in practical application scene
In, need to participate in the secret protection identification of face feature vector using the secret protection scheme of identified subject face image.
Being identified the secret protection of object facial image includes:
S201 uses the camera shooting identified object with identified object facial image privacy protection function
Facial photo；
S202 calculates the original feature vector M of the identified subject faceIFace；
S203, to the original feature vector M of the identified subject faceIFaceCarry out plus make an uproar processing, calculates identified pair
Add the feature vector M that makes an uproar as faceInoiseFace；Including using adaptive Gauss that device of making an uproar is added to carry out adding processing of making an uproar, the Gauss
Adding device of making an uproar includes the noise coefficient matrix M of pre-set N*NIk, default N-dimensional variance vectors MIσ_default, shooting people
The N-dimensional variance vectors M that face data statistics goes outIσ_statWith the MIkDiagonal line on N number of noise figure kIn；When the face number of shooting
When mesh < Ismall_num, MIσ=0.8*MIσ_default+0.2*
MIσ_stat；As the face of shooting number>Ismall_num and<Ibig_num, MIσ=0.5*MIσ_default+0.5*
MIσ_stat；The M as the face of shooting number > Ibig_numIσ=0.2*MIσ_default+0.8*MIσ_stat；When the face number of shooting
M when > > Ibig_num (the face number > 10*Ibig_num shot)Iσ=MIσ_stat；The Gauss adds the device Gauss that makes an uproar to add device of making an uproar
Every wheel generates the noise vector M of one group of N-dimensionalINoise=Gaussian Profile (0, MIk*MIσ), by the MIFaceValue and institute on often one-dimensional
State MINoiseIt is added, generate plus make an uproar feature vector MINoiseFace。
Carrying out identification to the identified object in the step S30 includes:
S301, the identified subject face add the feature vector M that makes an uproarINoiseFaceWith the references object face plus make an uproar
Feature vector MRNoiseFaceIt is compared；
S302 exports the face matching result of the identified object.
S301 is specifically included:
The identified subject face is added the feature vector M that makes an uproar by S3010INoiseFaceWith the references object face
Add the feature vector M that makes an uproarRNoiseFaceInput multilevel iudge device；
S3011, multilevel iudge device utilize SVM (Support Vector Machine) or Euclidean distance threshold value, judge institute
That states identified subject face adds the feature vector M that makes an uproarINoiseFaceWith the references object face respectively clustered in the database
Plus make an uproar feature vector MRNoiseFaceSimilarity degree.
As shown in figure 3, far call interface is disposed in the equipment in office for how having camera and having certain computing capability, such as
Intelligent monitoring camera, smart phone, PC etc..The far call interface is based on HTTP service or RPC service.For quilt
It identifies the interception of face in object video frame, is based on MTCNN or Mask RCNN framework.Multilevel iudge device application SVM supporting vector
Machine is directly compared using Euclidean distance.Adaptive Gauss adds the Rsmall_num used in device of making an uproar, Ismall_num with
Rbig_num, Ibig_num take 100,100 and 1000,1000 respectively.
The preferred embodiment of the present invention has been described in detail above.It should be appreciated that the ordinary skill of this field is without wound
The property made labour, which according to the present invention can conceive, makes many modifications and variations.Therefore, all technician in the art
Pass through the available technology of logical analysis, reasoning, or a limited experiment on the basis of existing technology under this invention's idea
Scheme, all should be within the scope of protection determined by the claims.
Claims (10)
1. a kind of face feature vector secret protection recognition methods, which comprises the steps of:
S10, references object facial image secret protection；
S20 is identified the secret protection of object facial image；
S30 identifies the identified object.
2. face feature vector secret protection recognition methods as described in claim 1, which is characterized in that the references object people
Face image secret protection includes:
S101 shoots the references object face using the camera with references object facial image privacy protection function and shines
Piece；
S102 calculates the original feature vector M of references object face described in network query function by face characteristicRFace；
S103, to the original feature vector M of the references object faceRFaceCarry out plus make an uproar processing, calculates the references object face
Portion adds the feature vector M that makes an uproarRNoiseFace；
The references object face is added the feature vector M that makes an uproar by S104RNoiseFaceIt is saved in database.
3. face feature vector secret protection recognition methods as claimed in claim 2, which is characterized in that the step S104 packet
It includes:
S1040 utilizes far call interface call service end；
The references object face is added the feature vector M that makes an uproar by S1041RNoiseFaceIt is transferred to the server-side；
S1042, the server-side add the feature vector M that makes an uproar for the references object faceRNoiseFaceIt is stored in database profession and goes forward side by side
Row cluster.
4. face feature vector secret protection recognition methods as claimed in claim 3, which is characterized in that the identified object
Facial image secret protection includes:
S201 uses the camera shooting identified subject face with identified object facial image privacy protection function
Photo；
S202 calculates the original feature vector M that subject face is identified described in network query function by face characteristicIFace；
S203, to the original feature vector M of the identified subject faceIFaceCarry out plus make an uproar processing, calculates described identified pair
Add the feature vector M that makes an uproar as faceINoiseFace。
5. face feature vector secret protection recognition methods as claimed in claim 4, which is characterized in that described to be known to described
Other subject face carries out identification
S301, the identified subject face add the feature vector M that makes an uproarINoiseFaceWith the references object face plus make an uproar feature
Vector MRNoiseFaceIt is compared；
S302 exports the face matching result of the identified object.
6. face feature vector secret protection recognition methods as claimed in claim 5, which is characterized in that the step S103 makes
Device of making an uproar is added to carry out adding processing of making an uproar with adaptive Gauss, the Gauss adds device of making an uproar to include the noise coefficient of pre-set N*N
Matrix MRk, default N-dimensional variance vectors MRσ_default, shooting the N-dimensional variance vectors M that counts of human face dataRσ_statAnd institute
State MRkDiagonal line on N number of noise figure kRn；As the face of shooting number < Rsmall_num, MRσ=0.8*MRσ_default+
0.2*MRσ_stat；As the face of shooting number>Rsmall_num and<Rbig_num, MRσ=0.5*MRσ_default+0.5*
MRσ_stat；The M as the face of shooting number > Rbig_numRσ=0.2*MRσ_default+0.8*MRσ_stat；When the face number of shooting
M when > > Rbig_numRσ=MRσ_stat。
7. face feature vector secret protection recognition methods as claimed in claim 6, which is characterized in that the Gauss adds device of making an uproar
Gauss adds the noise vector M of device of making an uproar one group of N-dimensional of every wheel generationRNoise=Gaussian Profile (0, MRk*MRσ), by the MRFacePer one-dimensional
On value and the MRNoiseIt is added, generate plus make an uproar feature vector MRNoiseFace。
8. face feature vector secret protection recognition methods as claimed in claim 7, which is characterized in that the step S203 packet
It includes and device of making an uproar is added to carry out adding processing of making an uproar using adaptive Gauss, the Gauss adds device of making an uproar to include the noise of pre-set N*N
Coefficient matrix MIk, default N-dimensional variance vectors MIσ_default, shooting the N-dimensional variance vectors M that counts of human face dataIσ_stat
With the MIkDiagonal line on N number of noise figure kIn；As the face of shooting number < Ismall_num, MIσ=0.8*
MIσ_default+0.2*MIσ_stat；As the face of shooting number>Ismall_num and<Ibig_num, MIσ=0.5*MIσ_default+
0.5*MIσ_stat；The M as the face of shooting number > Ibig_numIσ=0.2*MIσ_default+0.8*MIσ_stat；When the face of shooting
M when number > > Ibig_numIσ=MIσ_stat。
9. face feature vector secret protection recognition methods as claimed in claim 8, which is characterized in that the Gauss adds device of making an uproar
Gauss adds the noise vector M of device of making an uproar one group of N-dimensional of every wheel generationINoise=Gaussian Profile (0, MIk*MIσ), by the MIFacePer one-dimensional
On value and the MINoiseIt is added, generate plus make an uproar feature vector MINoiseFace。
10. face feature vector secret protection recognition methods as claimed in claim 9, which is characterized in that the step S301
Include:
The identified subject face is added the feature vector M that makes an uproar by S3010INoiseFaceWith the references object face plus make an uproar spy
Levy vector MRNoiseFaceInput multilevel iudge device；
S3011, the multilevel iudge device utilize SVM (Support Vector Machine) or Euclidean distance threshold value, judge institute
That states identified subject face adds the feature vector M that makes an uproarINoiseFaceWith the references object face respectively clustered in the database
Plus make an uproar feature vector MRNoiseFaceSimilarity degree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811640575.4A CN109753921A (en) | 2018-12-29 | 2018-12-29 | A Face Feature Vector Privacy-Preserving Recognition Method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811640575.4A CN109753921A (en) | 2018-12-29 | 2018-12-29 | A Face Feature Vector Privacy-Preserving Recognition Method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109753921A true CN109753921A (en) | 2019-05-14 |
Family
ID=66404475
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811640575.4A Pending CN109753921A (en) | 2018-12-29 | 2018-12-29 | A Face Feature Vector Privacy-Preserving Recognition Method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109753921A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110442683A (en) * | 2019-08-13 | 2019-11-12 | 北京明略软件系统有限公司 | The processing method and processing device of text information, storage medium, electronic device |
| CN110598464A (en) * | 2019-10-10 | 2019-12-20 | 山东浪潮人工智能研究院有限公司 | Data and model safety protection method of face recognition system |
| CN110648289A (en) * | 2019-08-29 | 2020-01-03 | 腾讯科技（深圳）有限公司 | Image denoising processing method and device |
| CN111274602A (en) * | 2020-01-15 | 2020-06-12 | 腾讯科技（深圳）有限公司 | Image characteristic information replacement method, device, equipment and medium |
| CN111401273A (en) * | 2020-03-19 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | A user feature extraction system and device for privacy protection |
| CN111414856A (en) * | 2020-03-19 | 2020-07-14 | 支付宝(杭州)信息技术有限公司 | Face image generation method and device for realizing user privacy protection |
| CN114140853A (en) * | 2021-12-03 | 2022-03-04 | 贵阳天依惠禾大数据有限公司 | Image desensitization and recognition method, device, computer equipment and readable storage medium |
| CN115211074A (en) * | 2021-02-10 | 2022-10-18 | 支付宝实验室(新加坡)有限公司 | Method and system for processing reference face |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060161527A1 (en) * | 2005-01-18 | 2006-07-20 | Cynthia Dwork | Preserving privacy when statistically analyzing a large database |
| WO2012046907A1 (en) * | 2010-10-05 | 2012-04-12 | 한국인터넷진흥원 | Apparatus for processing an image in a manner which protects privacy, image security system using same, and method therefor |
| WO2015026386A1 (en) * | 2013-08-19 | 2015-02-26 | Thomson Licensing | Method and apparatus for utility-aware privacy preserving mapping through additive noise |
| US20160210463A1 (en) * | 2012-08-20 | 2016-07-21 | Nadia Fawaz | Method and apparatus for utility-aware privacy preserving mapping through additive noise |
| CN105956518A (en) * | 2016-04-21 | 2016-09-21 | 腾讯科技（深圳）有限公司 | Face identification method, device and system |
| JP2017033372A (en) * | 2015-08-04 | 2017-02-09 | 日本放送協会 | Person recognition device and program thereof |
| CN107368752A (en) * | 2017-07-25 | 2017-11-21 | 北京工商大学 | A kind of depth difference method for secret protection based on production confrontation network |
| CN108446680A (en) * | 2018-05-07 | 2018-08-24 | 西安电子科技大学 | A kind of method for secret protection in face authentication system based on edge calculations |
-
2018
- 2018-12-29 CN CN201811640575.4A patent/CN109753921A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060161527A1 (en) * | 2005-01-18 | 2006-07-20 | Cynthia Dwork | Preserving privacy when statistically analyzing a large database |
| WO2012046907A1 (en) * | 2010-10-05 | 2012-04-12 | 한국인터넷진흥원 | Apparatus for processing an image in a manner which protects privacy, image security system using same, and method therefor |
| US20160210463A1 (en) * | 2012-08-20 | 2016-07-21 | Nadia Fawaz | Method and apparatus for utility-aware privacy preserving mapping through additive noise |
| WO2015026386A1 (en) * | 2013-08-19 | 2015-02-26 | Thomson Licensing | Method and apparatus for utility-aware privacy preserving mapping through additive noise |
| JP2017033372A (en) * | 2015-08-04 | 2017-02-09 | 日本放送協会 | Person recognition device and program thereof |
| CN105956518A (en) * | 2016-04-21 | 2016-09-21 | 腾讯科技（深圳）有限公司 | Face identification method, device and system |
| CN107368752A (en) * | 2017-07-25 | 2017-11-21 | 北京工商大学 | A kind of depth difference method for secret protection based on production confrontation network |
| CN108446680A (en) * | 2018-05-07 | 2018-08-24 | 西安电子科技大学 | A kind of method for secret protection in face authentication system based on edge calculations |
Non-Patent Citations (4)
| Title |
|---|
| JIE GUO等: "Person Re-Identification by Weighted Integration of Sparse and Collaborative Representation", 《IEEE ACCESS》 * |
| 侯小毛等: "云环境中考虑隐私保护的人脸图像识别", 《沈阳工业大学学报》 * |
| 巩道福等: "一种基于图像边缘的鲁棒水印算法", 《中国科学:信息科学》 * |
| 蒋金刚等: "基于人脸识别的Android手机隐私文本加密软件设计", 《网络安全技术与应用》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110442683A (en) * | 2019-08-13 | 2019-11-12 | 北京明略软件系统有限公司 | The processing method and processing device of text information, storage medium, electronic device |
| CN110648289A (en) * | 2019-08-29 | 2020-01-03 | 腾讯科技（深圳）有限公司 | Image denoising processing method and device |
| CN110648289B (en) * | 2019-08-29 | 2023-07-11 | 腾讯科技（深圳）有限公司 | Image noise adding processing method and device |
| CN110598464B (en) * | 2019-10-10 | 2021-04-27 | 浪潮集团有限公司 | A data and model security protection method for face recognition system |
| CN110598464A (en) * | 2019-10-10 | 2019-12-20 | 山东浪潮人工智能研究院有限公司 | Data and model safety protection method of face recognition system |
| CN111274602A (en) * | 2020-01-15 | 2020-06-12 | 腾讯科技（深圳）有限公司 | Image characteristic information replacement method, device, equipment and medium |
| CN111274602B (en) * | 2020-01-15 | 2022-11-18 | 腾讯科技（深圳）有限公司 | Image characteristic information replacement method, device, equipment and medium |
| CN111414856A (en) * | 2020-03-19 | 2020-07-14 | 支付宝(杭州)信息技术有限公司 | Face image generation method and device for realizing user privacy protection |
| CN111414856B (en) * | 2020-03-19 | 2022-04-12 | 支付宝(杭州)信息技术有限公司 | Face image generation method and device for realizing user privacy protection |
| CN111401273B (en) * | 2020-03-19 | 2022-04-29 | 支付宝(杭州)信息技术有限公司 | A user feature extraction system and device for privacy protection |
| CN111401273A (en) * | 2020-03-19 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | A user feature extraction system and device for privacy protection |
| CN115211074A (en) * | 2021-02-10 | 2022-10-18 | 支付宝实验室(新加坡)有限公司 | Method and system for processing reference face |
| CN114140853A (en) * | 2021-12-03 | 2022-03-04 | 贵阳天依惠禾大数据有限公司 | Image desensitization and recognition method, device, computer equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109753921A (en) | A Face Feature Vector Privacy-Preserving Recognition Method | |
| US11288504B2 (en) | Iris liveness detection for mobile devices | |
| Almeida et al. | Detecting face presentation attacks in mobile devices with a patch-based CNN and a sensor-aware loss function | |
| De Marsico et al. | Firme: Face and iris recognition for mobile engagement | |
| CN108197586B (en) | Face recognition method and device | |
| US10235603B2 (en) | Method, device and computer-readable medium for sensitive picture recognition | |
| Sun et al. | Photo assessment based on computational visual attention model | |
| CN104021380B (en) | Method and device for face recognition through computing device | |
| CN109800744A (en) | Image clustering method and device, electronic equipment and storage medium | |
| US10181073B2 (en) | Technologies for efficient identity recognition based on skin features | |
| US20230091865A1 (en) | Face image verification method and apparatus, electronic device, and storage medium | |
| Thavalengal et al. | User authentication on smartphones: Focusing on iris biometrics | |
| CN109086742A (en) | scene recognition method, scene recognition device and mobile terminal | |
| CN109886239B (en) | Portrait clustering method, device and system | |
| Zhang et al. | Privacy-friendly photo capturing and sharing system | |
| CN108734106A (en) | Quick sudden and violent probably video frequency identifying method based on comparison | |
| CN107294915A (en) | Image-pickup method, system and terminal | |
| WO2023000792A1 (en) | Methods and apparatuses for constructing living body identification model and for living body identification, device and medium | |
| Li et al. | Forensic analysis of digital images over smart devices and online social networks | |
| CN107729886B (en) | Method and device for processing face image | |
| CN114387674A (en) | Liveness detection method, system, device, storage medium and program product | |
| Sun et al. | iRyP: a purely edge-based visual privacy-respecting system for mobile cameras | |
| US10095915B2 (en) | Photo subscription system and method using biometric identification | |
| Dong et al. | Content-adaptive noise estimation for color images with cross-channel noise modeling | |
| Ali et al. | Iris recognition system in smartphones using light version (LV) recognition algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190514 |
|
| RJ01 | Rejection of invention patent application after publication |