• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    computer-implemented systems and methods for combining an encrypted biometric entry record with at least one stored encrypted biometric record, and without decryption of input data and at least one stored record.
    公开号:BR112019023750A2
    申请号:R112019023750-0
    申请日:2018-05-07
    公开日:2020-05-26
    发明作者:Streit Scott
    申请人:Veridium Ip Limited；
    IPC主号:
    专利说明:

    Invention Patent Descriptive Report for: SYSTEM AND METHOD FOR BIOMETRIC IDENTIFICATION
    FIELD OF THE INVENTION
    [001] The present invention relates, in general, to systems and methods to acquire and characterize biometric characteristics and, in particular, to systems and methods to acquire and characterize biometric characteristics in order to identify or authenticate a user.
    BACKGROUND OF THE INVENTION
    [002] Information of all types continues to be stored and accessed remotely, as in storage devices accessible by data communication networks. For example, many people and businesses store and access financial information, medical and health information, goods and services information, purchase information, entertainment information, multimedia information over the Internet or another communication network. In addition to accessing information, users can make money transfers (for example, purchases, transfers, sales or similar). In a typical scenario, a user registers to access information and then submits a username and password to log in and access the information. Protect access to (and from) that information and
    Petition 870200002163, of 1/6/2020, p. 6/56
    2/41 data stored in a data / communication network remains a primary concern.
    [003] Convenience directs consumers to biometrics-based access management solutions. The majority of smartphone users are believed to prefer using fingerprints instead of a password, with many prefering eye recognition over fingerprint recognition. Increasingly, biometrics is becoming a preferred and convenient method for identity detection and verification and authentication.
    [004] Transport-level encryption technology provides relatively strong protection for the transmission of various types of data, including biometric data, and supports confidentiality, security and non-repudiation requirements. Standards, such as IEEE 2410-2016, provide protection against an adversary who overhears communication and provide detailed mechanisms for authentication based on a pre-registered device and a previous identity, including the storage of a biometrics in encrypted form. This is an individual case and includes steps for sending and receiving encrypted biometric data, compared to an existing encrypted biometric sample. Therefore, this individual case is considered an authentication use case, as a given biometric vector and an identity can
    Petition 870200002163, of 1/6/2020, p. 7/56
    3/41 can be used as input and authentication can occur when the biometric vector corresponds to an existing biometric vector corresponding to the respective identity.
    SUMMARY
    [005] In one or more implementations, the present application provides systems and methods implemented by computer to combine an encrypted biometric entry record with at least one stored encrypted biometric record and without decryption of input data and at least one stored record. An initial biometric vector is supplied to a neural network, and the neural network converts the initial biometric vector into a vector of Euclidean measurable characteristic. The Euclidean measurable characteristic vector is stored in a store with other Euclidean measurable characteristic vectors. In addition, a current biometric vector representing the encrypted biometric input record is received from a mobile computing device over a data communication network, and the current biometric vector is supplied to the neural network. The neural network converts the current biometric vector into a current Euclidean measurable characteristic vector. In addition, a search for at least some of the stored Euclidean measurable characteristic vectors is performed on a portion of the data store using the current Euclidean measurable characteristic vector. The check-in
    Petition 870200002163, of 1/6/2020, p. 8/56
    4/41 encrypted biometric is matched with at least one encrypted biometric record in the encrypted space as a function of an absolute distance calculated between the current Euclidean measurable characteristic vector and a calculation of each of the respective Euclidean measurable characteristic vectors in the storage part.
    [006] In one or more implementations, the present application further provides the classification of the Euclidean measurable characteristic vector and / or the classification of the current Euclidean measurable characteristic vector, in which the classification is carried out at least in part using one or more functions from distance.
    [007] In one or more implementations, the classification of the Euclidean measurable resource and / or the current Euclidean measurable resource vector returns floating point values, and a Frobenius algorithm is used to calculate an absolute distance between each floating point and its mean .
    [008] In one or more implementations, the search is carried out in the Order log (n) time using the Frobenius algorithm to classify the Euclidean measurable biometric vectors that cross a hierarchy of the Euclidean measurable biometric vectors classified in the log (n) time Request; and identify that a respective vector
    Petition 870200002163, of 1/6/2020, p. 9/56
    5/41 Euclidean measurable biometric is the current Euclidean measurable characteristic vector.
    [009] In one or more implementations, the present application provides for the identification, for each respective Euclidean measurable biometric vector, of a plurality of floating point values and the use of a bitmap to eliminate from any calculation of absolute distances any plurality of values that are not present in all vectors.
    [0010] In one or more implementations, this application provides the identification, for each respective Euclidean measurable biometric vector, of a plurality of floating point values and the use of a bitmap to eliminate from any calculation of absolute distances any plurality of values that are not present in all vectors.
    [0011] In one or more implementations, this application provides to identify, for each respective Euclidean measurable biometric vector, a plurality of floating point values; and defining a sliding scale of importance based on the number of vectors, a respective of the floating point value appears.
    [0012] In one or more implementations, the neural network is configured with a variety of convolutional layers, together with a rectifier (ReLU) and cluster nodes.
    [0013] In one or more implementations, the neural network is configured to use clustering as a way of
    Petition 870200002163, of 1/6/2020, p. 10/56
    6/41 nonlinear sampling, and furthermore, in which one or more cluster nodes progressively reduce the spatial size of a Euclidean measurable characteristic vector represented to reduce the amount of parameters and computation in the neural network.
    [0014] In one or more implementations, the present application provides to identify the computation, for each of a plurality of measured Euclidean characteristic vectors, a relative position difference between a medium face vector and the respective measurable characteristic vector Euclidean, to square the difference in relative position, to add the values and to calculate the square root.
    [0015] In one or more implementations, the performance of the neural network is determined as a function of a cost function, in which a number of layers given as a spatial size of an output volume is calculated as a function of a volume size input W, a kernel field size (kernel field size) of layer K neurons, a step with which layers are applied S and a zero fill amount P used at an edge.
    [0016] These and other aspects, characteristics and advantages can be appreciated from the attached description of certain modalities of the invention, the attached figures and claims.
    Petition 870200002163, of 1/6/2020, p. 11/56
    7/41
    DESCRIPTION OF THE DRAWING FIGURES
    [0017] Figure 1 illustrates an exemplary system for identifying a user according to one or more modalities;
    [0018] Figure 2A is a block diagram that illustrates components and features of an example user computing device and includes various hardware and software components that serve to enable the operation of the system;
    [0019] Figure 2B illustrates a plurality of examples of modules, such as those that are coded in a storage and / or in memory, according to one or more modalities;
    [0020] Figure 2C is a block diagram that illustrates an exemplary configuration of a system server.
    [0021] Figure 3 is a system diagram that illustrates various aspects of the health monitoring and tracking system according to one or more modalities;
    [0022] Figures 4A and 4B show an example of an example of neural network in operation, according to an example of implementation of the present application;
    [0023] Figure 5 illustrates an example of a process according to the neural network; and
    [0024] Figure 6 is a flow chart that illustrates the steps of the example process according to an implementation.
    DETAILED DESCRIPTION
    [0025] Encryption remains a widely popular and effective way to protect information
    Petition 870200002163, of 1/6/2020, p. 12/56
    8/41 during the transportation of information. The nature of the information generally determines the level and type of encryption used to protect the information, especially to prevent compromises while the information is in transit. Unfortunately, it was not possible or practical to encrypt all data stored at the order level, for example, due to the need to search the data. It was impracticable, at least from a performance point of view, to search encrypted data effectively, without the need for an exhaustive search process that includes decrypting the data, record by record.
    [0026] Personally identifiable information (PII), in particular, requires encryption mechanisms and additional data protection policies and processes, as various operations on the data require decryption of that data for viewing and editing. For example, the Health Insurance Portability and Liability Act (HIPAA) requires data encryption during transport and offers policies for the release and dissemination of that data. The cryptographic strength policy aims to protect against compromises of PII databases, such as in cases of theft. When performing operations such as searching, without decryption, the data need not be exposed to possible compromises. Biometric data requires
    Petition 870200002163, of 1/6/2020, p. 13/56
    9/41 additional protections by processes and policies that introduce additional mechanisms, including more sophisticated encryption schemes, such as visual encryption.
    [0027] This application includes features and functionality to provide encrypted search, in addition to matching records one by one, given a reference biometry and a new entry record. In addition, the present application provides new methods and systems for searching encrypted biometric data without requiring data decryption on storage media, such as databases, file systems or other persistence mechanisms. In addition to a one-to-one implementation, one-to-many implementations are supported by the teachings contained herein, in which the search for encrypted biometric records can take place as a result of a newly received biometric record. In this case, an exhaustive search of the orthogonal group O (n) can be performed, in which each record is decrypted and compared. According to one or more implementations of the present application, an O (log n) solution is provided that does not require decryption and supports the location of a record without decryption. Generally referred to here as an identity use case, a given biometric vector is provided as input, and
    Petition 870200002163, of 1/6/2020, p. 14/56
    10/41 a respective biometric can be searched to determine if the biometric is in a database.
    [0028] In one or more implementations, this application provides a solution based on polynomials for identification in a large biometric encrypted data store, providing a system and mechanism to protect privacy. In addition, a selection of an initial biometry with low entropy is made. After that, a selection of a data structure provides an Order log (n) for a search. A selection of an algorithm for intermediate nodes is made later, so that the biometric cannot be discovered or a hashing algorithm cannot be reverted to the original biometric. Thus, an implementation according to the present application provides an end-to-end technique for biometrics to be used to provide identification through a database of thousands (or more) of individuals.
    [0029] In one or more implementations, a neural network, which may include a convolutional neural network, is used to process images and determine a correct cost function, such as to represent the performance quality of the neural network. In one or more implementations, a neural network is used to process other data, including audio content, such as a recording or representation of a person's voice. Although many implementations of
    Petition 870200002163, of 1/6/2020, p. 15/56
    11/41 examples shown and described in this document are related to the processing of one or more image files, a person skilled in the art will recognize that the present application is preferably biometric agnostic, and that any suitable biometric can be used in accordance with the teachings in this document. . Various types of neural networks are suitable for receiving various information formats and generating a vector of resources, such as a convolutional neural network, a recurrent neural network (RNN) or deep machine learning system.
    [0030] One or more implementations, a neural network, which may include a convolutional neural network, is used to process images and determine a correct cost function, such as to represent the performance quality of the neural network. In one or more implementations, a neural network is used to process other data, including audio content, such as a recording or representation of a person's voice. Although many of the example implementations shown and described in this document are related to the processing of one or more image devices, a person skilled in the art will recognize that the present application is preferably biometric agnostic and that any suitable biometric can be used in accordance with the teachings of this application. document. Other types of neural networks are suitable for receiving various information formats and
    Petition 870200002163, of 1/6/2020, p. 16/56
    12/41 generate a vector of resources, such as a recurrent neural network (RNN) or deep machine learning.
    [0031] An exemplary system for identifying a user is shown as a block diagram in Figure 1, which can be configured to interact with a neural network (not shown). In an exemplary arrangement, a system server 105, a remote computing device 102 and user computing devices 101a and 101b can be included. System 105 server can be virtually any computing device and / or data processing device capable of communicating with devices 101a, 101b and other remote computing devices 102, including receiving, transmitting and storing electronic information and requests process, as described hereinafter. The system server 105, remote computing device 102 and user devices 101a and 101b are intended to represent various forms of computers, such as laptops, desktop computers, computer workstations, personal digital assistants, servers, blade servers, mainframes and other computers and / or networked or cloud-based computing systems.
    [0032] In one or more implementations, the remote computing device 102 can be associated with a corporate organization, for example, a financial institution, a
    Petition 870200002163, of 1/6/2020, p. 17/56
    13/41 insurance company or any entity that maintains corporate user accounts (also known as transaction accounts). These corporate organizations provide services to account holders and require user authentication before granting access to corporate systems and services. As an additional example, remote computing device 102 may include payment networks and / or banking networks for processing financial transactions, as understood by those skilled in the art.
    [0033] User computing devices 101a, 101b can be any computing device and / or data processing device capable of incorporating the systems and / or methods described in this document, including but not limited to a personal computer, tablet , personal digital assistant, mobile electronic device, cell phone or smart phone device and the like. Transaction terminal 101b is intended to represent various forms of computing devices, such as workstations, dedicated point-of-sale systems, ATM terminals, personal computers, laptops, tablets, smart phone devices, personal digital assistants or others appropriate computers that can be used to conduct electronic transactions. Devices 101a, 101b can also be configured to receive user input, as well as capture and process
    Petition 870200002163, of 1/6/2020, p. 18/56
    14/41 biometric information, for example, digital images of a user, as described hereinafter.
    [0034] In one or more implementations, the system server 105 implements rules that govern access to information and / or the transmission of information between computing devices with which users interact (for example, device 101a, 101b) and a or more trusted back-end servers (for example, remote computing device 102).
    [0035] As described here, systems and methods for identifying and / or authenticating a user can meet the security levels required by a corporate system, using an API to integrate with an existing system (for example, the processing system transactions and data management of a financial institution). Therefore, the system server 105 does not need to know whether the underlying system (for example, remote computing device 102) is a Relational Database Management System (RDBMS), a Search Engine, a transaction processing system financial and similar. Consequently, systems and methods to facilitate secure authentication offer a point-and-cut mechanism to add appropriate security to existing corporate systems, as well as systems under development. In some implementations, the architecture of
    Petition 870200002163, of 1/6/2020, p. 19/56
    15/41 system is a language neutral, allowing REST, JSON and Secure Socket Layers to provide the communication interface between the various computing devices (for example, 101a, 101b, 102 and 105). In addition, in one or more implementations, the architecture is built with the servlet specification, open secure socket layers, Java, JSON, REST and / or Apache Solr. Therefore, the systems disclosed to authenticate a user can implement open standards, thus allowing significant interoperability.
    [0036] It should also be understood that, while the various computing devices and machines mentioned herein, including, but not limited to, user devices 101a and 101b, system server 105 and remote computing device 102 are referred to herein as individual / unique devices / machines. In certain implementations, the referenced devices and machines, and their associated and / or tracked operations, resources and / or functionality, can be combined or organized or otherwise employed on any number of devices and / or machines, such as in a connection network or wired connection, as is known to those skilled in the art.
    [0037] Figure 2A is a block diagram that illustrates components and features of a user computing device 101a, and includes several hardware components and
    Petition 870200002163, of 1/6/2020, p. 20/56
    16/41 software used to enable the operation of the system, including one or more processors 110, memory 120, microphone 125, display 140, camera 145, audio output 155, storage 190 and a communication interface 150. Processor 110 serves to run a client application in the form of software instructions that can be loaded into memory 120. Processor 110 can be a number of processors, a central processing unit CPU, a graphics processing unit GPU , a multiprocessor kernel or any other type of processor, depending on the specific implementation.
    [0038] Preferably, memory 120 and / or storage 190 are accessible by processor 110, thus allowing the processor to receive and execute instructions encoded in memory and / or storage, in order to make the device and its various hardware components perform operations for aspects of the exemplary systems and methods disclosed herein. The memory can be, for example, a random access memory (RAM) or any other suitable volatile or non-volatile computer-readable storage medium. Storage 190 can take various forms, depending on the specific implementation. For example, storage may contain one or more components or devices, such as a hard disk, flash memory, rewritable optical disc, rewritable magnetic tape or some
    Petition 870200002163, of 1/6/2020, p. 21/56
    17/41 combination of the above. In addition, memory and / or storage can be fixed or removable.
    [0039] One or more software modules 130 are encoded in storage 190 and / or memory 120. Software modules 130 may comprise one or more programs or software applications with computer program code or a set of instructions executed in the processor 110. As shown in Figure 2B, one or more of a user interface module 170, a biometric capture module 172, an analysis module 174, an enrollment module 176, a database module 178, a module authentication codes 180 and a communication module 182 can be included among software modules 130 that are executed by processor 110. Such computer program codes or instructions configure processor 110 to perform system operations and methods disclosed herein and can be written in any combination of one or more programming languages.
    [0040] The program code can be executed entirely on the computing device of user 101, as a standalone device, partially on the computing device of user 101, partially on the system server 105 or entirely on the system server 105 or on another computer / remote device 102. In the last scenario, the remote computer can be connected
    Petition 870200002163, of 1/6/2020, p. 22/56
    18/41 to the user 101 computing device over any type of network, including a local area network (LAN) or a wide area network (WAN), mobile communications network, cellular network or the connection can be made to an external computer (for example, over the Internet using an Internet service provider).
    [0041] It may also be that the program code of software modules 130 and one or more computer-readable storage devices (such as memory 120 and / or storage 190) form a computer program product that can be manufactured and / or distributed according to the present invention, as is known to those skilled in the art.
    [0042] It should be understood that, in some illustrative modalities, one or more of the software modules 130 can be downloaded over a network for storage 190 of another device or system via communication interface 150. In addition, it should be noted that other information and / or data relevant to the operation of the present systems and methods (such as database 185) can also be stored in storage. In some implementations, this information is stored in an encrypted data store that is specifically allocated to securely store information collected or generated by processor 110 running software modules 130.
    Petition 870200002163, of 1/6/2020, p. 23/56
    19/41
    Preferably, encryption measures are used to store the information locally in the user's computing device storage and the transmission of information to the system's 105 server. For example, this data can be encrypted using a 1024-bit polymorphic cipher or, depending on export controls, a 256-bit AES encryption method. In addition, encryption can be performed using remote key (seeds) or local keys (seeds). Alternative encryption methods can be used as would be understood by those skilled in the art, for example, SHA256.
    [0043] In addition, data stored on the user's 101a computing device and / or system 105 server can be encrypted using an encryption key generated from biometric information, vividness information or information from the user's computing device, as described here below. In some implementations, a combination of the precedent can be used to create a complex unique key for the user that can be encrypted on the user's computing device using Elliptical Curve Encryption, preferably at least 384 bits long. In addition, this key can be used to protect user data
    Petition 870200002163, of 1/6/2020, p. 24/56
    20/41 stored on the user's computing device and / or the system server.
    [0044] Also preferably stored in storage 190 is database 185. As will be described in more detail below, the database contains and / or maintains various items and data elements that are used throughout the various operations of the system and method to authenticate a user who conducts a financial process, transaction in a transaction terminal. The information stored in database 185 may include, without limitation, a user profile, as will be described in more detail here. It should be noted that, although database 185 is described as locally configured for user computing device 101a, in certain implementations, the database and / or several of the data elements stored therein can, additionally or alternatively, be located remotely (as on a remote device 102 or system server 105 - not shown) and connected to the user's computing device over a network in a manner known to those skilled in the art.
    [0045] A 115 user interface is also operationally connected to the processor. The interface can be one or more input or output devices, such as switch (es), button (s), key (s), touchscreen, microphone, etc., as would be understood in the art of electronic computing
    Petition 870200002163, of 1/6/2020, p. 25/56
    21/41 devices. User interface 115 serves to facilitate the capture of user commands, such as on-off commands or user information and settings related to the operation of the system for user authentication. For example, the interface serves to facilitate the capture of certain information from the user's computing device 101, such as the user's personal information to register with the system, in order to create a user profile.
    [0046] The computing device 101a may also include a monitor 140 that is also operationally connected to the processor 110 processor. The monitor includes a screen or any other presentation device that allows the user's computing device to instruct or provide feedback to the user about the operation of the system 100. As an example, the screen can be a digital screen, such as a dot matrix screen or another two-dimensional screen.
    [0047] As an example, the interface and the screen can be integrated with a touch screen. Consequently, the display is also used to show a graphical user interface, which can display various data and provide forms that include fields that allow information to be entered by the user. Touching the touchscreen in locations corresponding to the display of a graphical user interface allows the user to interact with the device
    Petition 870200002163, of 1/6/2020, p. 26/56
    22/41 to enter data, change settings, control functions, etc. Therefore, when the touch screen is touched, user interface 115 communicates this to processor 110 and settings can be changed or information entered by the user can be captured and stored in memory 120 and / or storage 190.
    [0048] Devices 101a, 101b may also include a camera 145 capable of capturing digital images. The camera may be one or more imaging devices configured to capture images of at least a part of the user's body, including the user's eyes and / or face while using the user's 101a computing device. The camera serves to facilitate the capture of user images for the purpose of image analysis by the processor of the configured user's computing device, which includes the identification of biometric resources to (biometrically) authenticate the user of the images. User computing device 101a and / or camera 145 may also include one or more light or signal emitters (not shown), for example, a visible light emitter and / or infrared light emitter and the like. The camera can be integrated with the user's computing device, such as a front or rear camera that incorporates a sensor, for example, and without limitation, a CCD or CMOS sensor. Alternatively, the camera may be external to the user's computing device 101a. At
    Petition 870200002163, of 1/6/2020, p. 27/56
    23/41 possible variations of the camera and light emitters would be understood by those skilled in the art. In addition, the user's computing device may also include one or more microphones 125 to capture audio recordings, as understood by those skilled in the art.
    [0049] An audio output 155 can also be operationally connected to processor 110. The audio output can be integrated into the user's computing device 101 or external to the user's computing device and can be any type of speaker system configured to play electronic audio files as would be understood by those skilled in the art.
    [0050] Several hardware devices / sensors 160 can be operationally connected to the processor. Sensors 160 may include: an on-board clock to track the time of day; a GPS-enabled device for determining the location of the user's computing device; an accelerometer to track the orientation and acceleration of the user's computing device; a gravity magnetometer to measure the earth's magnetic field, proximity sensors to measure the distance from the user's computing device to an object, RF radiation sensors to measure radiation and other devices to capture information about the computing device's environment of
    Petition 870200002163, of 1/6/2020, p. 28/56
    24/41 user, as would be understood by those skilled in the art.
    [0051] Communication interface 150 is also operationally connected to processor 110 and can be any interface that allows communication between the user's computing device 101a and external devices, machines and / or elements. Preferably, the communication interface may include, among others, a modem, a network interface card (NIC), an integrated network interface, a radio frequency transmitter / receiver (for example, Bluetooth, mobile, NFC), a satellite communication transmitter / receiver, an infrared port, a USB connection and / or any other interface to connect the user's computing device to other computing devices and / or communication networks, such as private networks and the Internet. Such connections may include a wired or wireless connection (for example, using the 802.11 standard), although it should be understood that the communication interface can be almost any interface that allows communication to / from the user's computing device.
    [0052] Figure 2C is a block diagram illustrating an exemplary configuration of the system server 105. The system server 105 may include a processor 210 that is operationally connected to various components of
    Petition 870200002163, of 1/6/2020, p. 29/56
    25/41 hardware and software that serve to allow the operation of the system to, for example, authenticate a user in connection with a transaction at a transaction terminal. Processor 210 serves to execute instructions to perform various operations, including those related to user authentication and transaction processing / authorization. Processor 210 may be a number of processors, a multiprocessor core or some other type of processor, depending on the specific implementation.
    [0053] In certain implementations, a memory 220 and / or storage medium 290 is accessible by processor 210, thus allowing processor 210 to receive and execute instructions stored in memory 220 and / or storage 290. Memory 220 can be , for example, a random access memory (RAM) or any other suitable volatile or non-volatile computer-readable storage medium. In addition, memory 220 can be fixed or removable. Storage 290 can take various forms, depending on the specific implementation. For example, storage 290 may contain one or more components or devices, such as hard disk, flash memory, rewritable optical disc, rewritable magnetic tape, or some combination of the above. Storage 290 can also be fixed or removable.
    Petition 870200002163, of 1/6/2020, p. 30/56
    26/41
    [0054] One or more software modules 130 (represented in Figure 2B) can be encoded in storage 290 and / or memory 220. Software modules 130 can comprise one or more software programs or applications (collectively referred to as secure authentication) server application) with computer program code or a set of instructions executed on processor 210. That computer program code or instructions for performing operations for aspects of the systems and methods disclosed herein can be written in any combination of one or more languages programming, as would be understood by those skilled in the art. The program code can be run entirely on the system server 105 as a standalone software package, partly on the system server 105 and partly on a remote computing device, such as a remote computing device 102, remote computing device 102, device user computing 101a and / or user computing device 101b, or entirely on those remote computing devices.
    [0055] Also preferably stored in storage 290 is a database 280. As will be described in more detail below, database 280 contains and / or maintains various items and data elements that are used throughout the various operations of the system 100, including but not limited to user profiles, as will be described in more
    Petition 870200002163, of 1/6/2020, p. 31/56
    27/41 details in this document. It should be noted that, although database 280 is represented as configured locally for computing device 105, in certain implementations, database 280 and / or several of the data elements stored on it can be stored in memory computer readable or on a storage medium that is located remotely and connected to the system 105 server via a network (not shown), in a manner known to those skilled in the art.
    [0056] A communication interface 250 is also operationally connected to processor 210. The communication interface can be any interface that allows communication between the system server 105 and devices, machines and / or external elements. In certain implementations, the communication interface may include, among others, a modem, a network card (NIC), an integrated network interface, a radio frequency transmitter / receiver (for example, Bluetooth, cellular, NEC), a communication satellite transmitter / receiver, an infrared port, a USB connection and / or any other interface to connect computing device 105 to other computing devices and / or communication networks, such as private networks and the Internet. Such connections may include a wired or wireless connection (for example, using the 802.11 standard), although it should be understood that the 255 communication interface can be
    Petition 870200002163, of 1/6/2020, p. 32/56
    28/41 virtually any interface that allows communication to / from processor 210.
    [0057] The operation of the system to authenticate a user and the various elements and components described above will be further appreciated with reference to Figures 3-4B and with continued reference to Figures 1 and 2A-2B. The processes represented in Figures 3 and 4 are shown from the perspective of the computing device of the user 101a, as well as of the system server 105, however, it must be understood that the processes can be executed, in whole or in part, by the device computing device 101a, server system 105 and / or other computing devices (e.g., remote computing device 102) or any combination thereof.
    [0058] Figure 3 is a simple diagram that illustrates a series of nodes XI, X2 and X3 in a neural network and also illustrates an activation function as a sum of a matrix of weights times the values X, and that generates a value Y. One goal is to find the ideal number of neurons and layers in the neural network. A formula for calculating the number of neurons that adjust to a given volume is given by (W - K + 2 P) / S + 1. For example, the cost function is Euclidean and the number of layers given as the spatial size of the output volume can be calculated as a function of the size of the input volume W, the size of the nucleus field of the K neurons
    Petition 870200002163, of 1/6/2020, p. 33/56
    29/41 of the Conv layer, the pitch with which S is applied and the amount of zero fill P used at the edge.
    [0059] For example, the neural network is initially trained as a classifier using labeled biometric data. As part of the training process, each person is stored and images of the training are present. After an initial biometric vector (IBV) is introduced into the neural network, the vector used in layer n-1 can be used as a unique feature vector to identify that initial biometric vector, albeit in the homomorphic encrypted space. This resource vector is measurable and encrypted Euclidean. In addition, the feature vector replaces the biometric vector. In one or more implementations, the resource vector is a list of 256 floating point numbers and the biometric is reconstructed from the list of 256 floating point numbers. The resource vector is therefore unidirectional encryption. The feature vectors have the property of being Euclidean measurable.
    [0060] It is recognized that the present application provides applicability in a variety of vertical markets that may require an encrypted search for identification using biometric entries. For example, insurance companies and banks want mechanisms to identify an account holder who has lost his account number. The solution
    Petition 870200002163, of 1/6/2020, p. 34/56
    30/41 provided in accordance with the present application can be executed in time O (log (n)), which is believed to improve the existing O (n) algorithms for encrypted research in a database of encrypted biometric records.
    [0061] In one or more implementations, for example, a mathematical transformation is employed to provide partitioning for a linear search. In this case, this transformation takes a biometric vector and returns a resource vector. The resource vector is later usable for direct access to a node. The accessed node can be checked linearly to determine a respective identity. For example, an individual user can be identified as a function of the node that is accessed as a function of the resource vector.
    [0062] In one or more implementations, a mathematical function can be used that partitions information (for example, vectors) associated with biometrics to support a tree structure, such as a B + tree. Alternatively, or in addition, the information associated with biometrics (for example, biometric vectors) can be placed through a neural network, which is operable to create a non-reversible biometric for correspondence or proximity.
    [0063] In connection with an implementation that employs a neural network, a Euclidean distance algorithm works
    Petition 870200002163, of 1/6/2020, p. 35/56
    31/41 to convert a facial biometric vector or other value into something that can be compared. An initial biometric value (model) is processed through the neural network and results in a characteristic vector. This implementation is useful, for example, in connection with a use case one for one or a case for many. In both cases, the neural network and the resource vector are usable to perform the comparison in the encrypted space. In addition, and to post an individual case, a neural network can operate to perform a Euclidean cost to access a specific biometry or one or more groups of biometrics, after applying various levels of the neural network. The resulting vector, after matrix multiplications for each respective layer, can use a Euclidean distance algorithm, based on the Euclidean cost function, and is referred to here, generally, as a characteristic vector.
    [0064] The computing system can include clients and servers. A client and a server are usually remote from each other and usually interact over a communication network. The relationship between the client and the server arises because of computer programs that run on the respective computers and have a client-server relationship.
    [0065] Figures 4A and 4B show an example of an example neural network in operation, according to an example of
    Petition 870200002163, of 1/6/2020, p. 36/56
    32/41 implementation of this application. The initial image is applied to the neural network. The resulting softmax is provided, which shows which of the defined buckets the image belongs to. In the respective bucket shown, images are provided for the individual in question. For example, the resource vector for person X, such as 256 values, when applied to the classification function results in a floating point number of Y. For all values close to Y, the individual can be identified, provided that the vector measurable resource is Euclidean and the function and classification is stable.
    [0066] Continuing with reference to Figures 4A and 4B, a resulting softmax function shows an image of the person in question in the training set. Subsequently, the matrix calculations for creating the softmax are shown. Two convolutional steps prior to softmax are fc7conv, which has the output of convolutional layer 7. The output is the characteristic vector.
    [0067] Figure 5 illustrates an example of a process according to the neural network. A variety of convolutional layers are provided, along with a rectifier (ReLU) and cluster nodes. In the example shown in Figure 5, the rectifier is an activation function suitable for deep neural networks and is an activation function defined as f (x) = max (0, x).
    Petition 870200002163, of 1/6/2020, p. 37/56
    33/41
    [0068] In addition to the ReLU, the neural network of this application can be configured to use the pool as a form of non-linear sampling. More particularly, a type of clustering algorithm, referred to here generally as maxpool, partitions input images into a set of non-overlapping rectangles and, for each sub-region, generates the maximum. Once a resource is found, its exact location may be less important than its relationship to other resources. The grouping layer function progressively reduces the spatial size of the representation to reduce the amount of parameters and computation on the network. Thus, through the use of convolutions, ReLU and maxpool, a vector of reasonable size of 256 is provided for use in correspondence, classification and comparison. In addition, the encrypted Euclidean resource vectors (EMFVs) resulting from the neural network can later be used in a classification algorithm to receive a scalar value that can be compared to resource vectors from other biometries. In one or more implementations, the present application uses pre-trained models, such as deep learning models trained in data associated with face images.
    [0069] Continuing with reference to Figure 5, the characteristic vector is an index for use in storage. For indexing, a classification algorithm is used which,
    Petition 870200002163, of 1/6/2020, p. 38/56
    34/41 given an input resource vector, it returns a floating point number, which can be used as an index for storage. The classification algorithm depends on a high quality medium vector, which helps to create distances between people. The classification algorithm allows you to search and find a person enrolled in polynomial time.
    [0070] The floating point number mentioned above allows an index search for a previously stored biometry. In addition, the present application employs a classification algorithm to provide the intended person based on the input resource vector.
    [0071] In one or more implementations, an algorithm is used in which data and matrices are stored to provide a learned model. In this case, a medium face vector is compared to a respective IBV, and the differences between the medium face vector and the IBV are calculated. For example, a Frobenius algorithm can be used in which an absolute distance is calculated between each floating point and its mean. In addition, the relative position difference between the medium face vector and the respective IBV is calculated and, subsequently, squared. All values are added together and the square root is calculated. This results in relatively close values, that is, all distances
    Petition 870200002163, of 1/6/2020, p. 39/56
    Frobenius' calculated 35/41 are relatively close groupings.
    [0072] In one or more implementations, during registration, a number of facial vectors (0-12) is obtained. The vectors are processed by the neural network and, if any of the processed resource vectors has a value of 0, the value will be degraded to (approach 0 or obtain 0), since the value is not significant because it is not present in all vector resources .
    [0073] In operation, every time an image is used, it is degraded, which provides more separation within the resource vectors. Given a characteristic vector of 256 non-negative integers, the procedure described here classifies the vector into an appropriate pool, and each pool is for one and only one distinct person. If the person is not found, it is assumed that the algorithm will search for neighboring people to complete the search.
    [0074] Thus, the present application uses Euclidean distance to classify unknown vectors of the form x = (x 0 , x 1 , ···, x 255 ). Unless otherwise stated, x t is a non-negative integer for all i G {0.1, ··, 255}. The distance between two vectors is defined as d (x, y) = (Σ 2 = οΙ * ί -yd 2 ) 1/2 · Given a vector x, let x = x / d (x, (0,0, · ·, 0)).
    Petition 870200002163, of 1/6/2020, p. 40/56
    36/41
    [0075] In operation, U is defined as the set of known vectors and | U | the number of vectors in U. | í / | 00 is assumed. Then the average of all vectors in U is given by m y = ι / l to 1 x and can be calculated explicitly if | í / | it is small enough. If | í / | is not small enough, so m y can be approximated. Note that the coordinates of m y are not necessarily non-negative integers.
    [007 6] Now, consider a partition of U as U = [J, P,. We observed experimentally that for each Pj there is ctj, bj G (0, 00) such that dÇmu.y) G [aj, bj] for all y G Pj. In addition, [aj, bj] is separated from [a k , b k ] for jtk. In other words, distance ranges of the average vector can be used to classify unknown vectors.
    [0077] Given a vector y ¢. U, is calculated to determine how to exclusively extend partitions of U to partitions of V = U U {y}. If d (mh, y) ¢. [aj, bj] for all j, then the interval [aj, bj] closest to y is chosen and the subset Pj associated with that interval is chosen to include y in the extension of the original partition to a partition of V. If it happens that y is equidistant to two different intervals, the subset in which to include y in the V partition is not well defined. In this case, the numerical results
    Petition 870200002163, of 1/6/2020, p. 41/56
    37/41 must be reexamined and a better choice for at least one of the two equidistant intervals of y must be made.
    [0078] In an example operation, 257 images are used for training. In the training operation example, the neural network is a network of 32 nodes and 8 layers of convolution. The training allocates the appropriate weights to the network. With this trained network, a new face was applied for storage or research. In both cases, the face was placed through the neural network and, in the convolutional layer 7, a characteristic vector was received.
    [0079] Using a set of images and applying our neural network, the following EMFV was received in the convolutional layer 7. An example of one of these outputs is below:
    30007 12 402200020200101347041 482026087500047032012340000002 002163021100408350531409031610 281020223104202521022001013211 019553010400421541051315817470 024144030040940031030101010101010101010101010101010
    [0080] The characteristic vector is classified for a specific person at run time O (1). Using the algorithm described here,
    Petition 870200002163, of 1/6/2020, p. 42/56
    38/41 three individuals. Using normalized vectors, the resulting image ranges are as follows. For Person 1, normalized distances range from 0.85 to 1.12. For Person 2, the range is 1.18 to 1.32. For Person 3, normalized distances range from 0.39 to 0.68.
    [0081] As a practical example, Person 1's subsequent IBVs, when supplied to the neural network and the classification algorithm, produce results between 0.85 and 1.12. Likewise, Person 3's subsequent IBVs produce results between 0.39 and 0.68. These acceptable ranges offer ranges for each respective person. Therefore, the idea is that bands and no collisions are observed for bands between people. For small data sets, the results are accurate.
    [0082] Thus, this application provides technology for:
    (1) acquire a biometric match, (2) plain text biometric match, (3) encrypt the biometric, (4) perform a measurable Euclidean match, and (5) search using an indexing scheme. They are provided in a manner protected by privacy and based on polynomial time, which is also agnostic and biometric and performed according to machine learning. The present application provides a general-purpose solution that produces measurable Euclidean biometric cipher text, including depending on the convolutional neural network. This is also provided
    Petition 870200002163, of 1/6/2020, p. 43/56
    39/41 due to a classification algorithm being used for one-to-many identification, which maximizes privacy and runs between 0 (1) and 0 (log (n)).
    [0083] An initial biometric vector (IBV) is received and one or more modules apply a neural network to the IBV. The IBV is processed, on the user's computing device or on a server, through a set of matrix operations, to create a resource vector. Any part or the entire resource vector can be stored on the client or on the server. After multiplying the matrix in several layers, the IBV is returned as a measurable Euclidean vector. In one or more implementations, the same action can occur with the current biometric vector (CBV) and the two vectors are matched.
    [0084] Figure 6 is a flowchart that illustrates the example 600 process steps according to an implementation. It should be considered that several of the logical operations described in this document are implemented (1) as a sequence of acts implemented by a computer or program modules running on a communication device and / or (2) as interconnected machine logic circuits or modules circuit within a communication device. Implementation is a matter of choice, depending on the requirements of the device (for example, size, energy, consumption, performance, etc.). Therefore, operations
    Petition 870200002163, of 1/6/2020, p. 44/56
    40/41 logics described in this document are referred to in various ways as operations, structural devices, acts or modules. Several of these operations, structural devices, acts and modules can be implemented in software, in firmware, in digital logic for special purposes and in any combination thereof. It should also be appreciated that more or less operations can be performed than those shown in the figures and described here. These operations can also be performed in a different order than described here. Example steps 600 illustrated in Figure 6 provide a computer-implemented method for combining an encrypted biometric entry record with at least one stored encrypted biometric record with no decryption of input data and at least one stored record.
    [0085] The process begins at step 602, and an initial biometric vector is supplied to a neural network, and the neural network converts the initial biometric vector into a vector of measurable Euclidean characteristic (step 604). The Euclidean measurable characteristic vector is stored in a store with other Euclidean measurable characteristic vectors (step 606). In addition, a current biometric vector representing the encrypted biometric entry record is received from a mobile computing device over a data communication network.
    Petition 870200002163, of 1/6/2020, p. 45/56
    41/41 and the current biometric vector is supplied to the neural network (step 608). The neural network converts the current biometric vector into a current Euclidean measurable characteristic vector (step 610). In addition, a search for at least some of the stored Euclidean measurable characteristic vectors is performed on a portion of the data store using the current Euclidean measurable characteristic vector (step 614). The encrypted biometric input record is matched with at least one encrypted biometric record in the encrypted space as a function of an absolute distance calculated between the current Euclidean measurable feature vector and a calculation of each of the respective Euclidean measurable feature vectors in the storage part (step 616).
    [0086] The subject described above is provided for illustrative purposes only and should not be construed as limiting. Various modifications and alterations can be made to the object described here without following the examples of modalities and applications illustrated and described, and without departing from the true spirit and scope of the present invention, including as stated in each of the following claims.
    权利要求:
    Claims (20)
    [1]
    1. Computer-implemented method to combine an encrypted biometric entry record with at least one stored encrypted biometric record, with no decryption of input data and at least one stored record, the method characterized by the fact that it comprises:
    supply an initial biometric vector to a neural network, where the neural network translates the initial biometric vector into a vector of Euclidean measurable characteristic;
    store the Euclidean measurable characteristic vector in storage with other Euclidean measurable characteristic vectors;
    classification of the Euclidean measurable characteristic vector; and / or classification of the current Euclidean measurable characteristic vector, in which the classification is performed at least in part using one or more distance functions;
    receive, from a mobile computing device on a data communication network, a current biometric vector representing the encrypted biometric input record;
    provide the current biometric vector for the neural network, where the neural network translates the current biometric vector into a current Euclidean measurable characteristic vector; and
    Petition 870200002745, of 07/01/2020, p. 6/17
    [2]
    2/10 conduct a search of at least some of the Euclidean measurable feature vectors stored in a portion of the data store using the current Euclidean measurable feature vector, where the encrypted biometric entry record is matched with at least one biometric record encrypted in the encrypted space as a function of an absolute distance calculated between the current Euclidean measurable characteristic vector and a calculation of each of the respective Euclidean measurable characteristic vectors in the storage portion, where the classification of the Euclidean measurable resource and / or the vector current Euclidean measurable resource returns floating point values, and a Frobenius algorithm is used to calculate an absolute distance between each floating point and its mean.
    2. Method, according to claim 1, characterized by the fact that the search is performed at the time of the Order log (n).
    [3]
    3. Method, according to claim 1, characterized by the fact that it still comprises:
    use a Frobenius algorithm to classify Euclidean measurable biometric vectors;
    Petition 870200002745, of 07/01/2020, p. 7/17
    3/10 go through a hierarchy of Euclidean measurable biometric vectors classified at the time of the Order log (n); and to identify that a respective Euclidean measurable biometric vector is the current Euclidean measurable characteristic vector.
    [4]
    4. Method, according to claim 1, characterized by the fact that it still comprises:
    identify, for each respective Euclidean measurable biometric vector, a plurality of floating point values; and use a bitmap to eliminate from a calculation of absolute distances any plurality of values that are not present in all vectors.
    [5]
    5. Method, according to claim 1, characterized by the fact that it still comprises:
    identify, for each respective Euclidean measurable biometric vector, a plurality of floating point values; and define a sliding scale of importance based on the number of vectors, a respective of the floating point value appears.
    [6]
    6. Method, according to claim 1, characterized by the fact that the neural network is configured
    Petition 870200002745, of 07/01/2020, p. 8/17
    4/10 with a variety of convolutional layers, together with a rectifier (ReLU) and cluster nodes.
    [7]
    7. Method, according to claim 1, characterized by the fact that the neural network is configured to use clustering as a form of non-linear sampling, and in which one or more cluster nodes progressively reduce the spatial size of a vector of measurable Euclidean characteristic represented to reduce the amount of parameters and computation in the neural network.
    [8]
    8. Method, according to claim 6, characterized by the fact that it still comprises:
    calculating, for each of a plurality of stored Euclidean measurable feature vectors, a relative position difference between a medium-faced vector and the respective Euclidean measurable feature vector;
    quadrature of the relative position difference;
    add the values; and calculate the square root.
    [9]
    9. Method, according to claim 1, characterized by the fact that the performance of the neural network is determined as a function of a cost function, in which a number of layers given as a spatial size of
    Petition 870200002745, of 07/01/2020, p. 9/17
    5/10 an output volume is computed as a function of an input volume size W, the size of the nucleus field of the K layer neurons, a step with which the layers are applied S and a zero fill amount P used on an edge.
    [10]
    10. Method, according to claim 1, characterized by the fact that the neural network translates the initial biometric vector and the current biometric vector as a function of matrix multiplications for each respective layer and uses a Euclidean distance algorithm based on a Euclidean cost function.
    [11]
    11. System implemented by computer to combine an encrypted biometric entry record with at least one stored encrypted biometric record, without decryption of input data and at least one stored record, the system characterized by the fact that it comprises:
    one or more processors and a computer-readable medium, where the one or more processors are configured to interact with the computer-readable medium to perform operations that include:
    supply an initial biometric vector to a neural network, where the neural network translates the initial biometric vector into a vector of Euclidean measurable characteristic;
    Petition 870200002745, of 07/01/2020, p. 10/17
    6/10 store the Euclidean measurable characteristic vector in a storage with other Euclidean measurable characteristic vectors;
    receive, from a mobile computing device on a data communication network, a current biometric vector representing the encrypted biometric input record;
    classification of the Euclidean measurable characteristic vector; and / or classification of the current Euclidean measurable characteristic vector, in which the classification is performed at least in part using one or more distance functions;
    provide the current biometric vector for the neural network, where the neural network translates the current biometric vector into a current Euclidean measurable characteristic vector; and conduct a search for at least some of the Euclidean measurable feature vectors stored in a portion of the data store using the current Euclidean measurable feature vector, where the encrypted biometric input record is matched with at least one encrypted biometric record in the encrypted space as a function of an absolute distance calculated between the current Euclidean measurable characteristic vector and a calculation of each of the respective Euclidean measurable characteristic vectors in the storage portion,
    Petition 870200002745, of 07/01/2020, p. 11/17
    7/10 where the classification of the Euclidean measurable resource and / or the current Euclidean measurable resource vector returns floating point values, and a Frobenius algorithm is used to calculate an absolute distance between each floating point and its mean.
    [12]
    12. System, according to claim 11, characterized by the fact that the search is performed at the time of the Order log (n).
    [13]
    13. System according to claim 11, characterized by the fact that the one or more processors are configured to interact with the computer-readable medium to perform operations that include:
    use a Frobenius algorithm to classify Euclidean measurable biometric vectors;
    go through a hierarchy of Euclidean measurable biometric vectors classified at the time of the Order log (n); and to identify that a respective Euclidean measurable biometric vector is the current Euclidean measurable characteristic vector.
    [14]
    14. System according to claim 11, characterized by the fact that the one or more processors are configured to interact with the computer-readable medium to perform operations that include:
    Petition 870200002745, of 07/01/2020, p. 12/17
    8/10 identify, for each respective Euclidean measurable biometric vector, a plurality of floating point values; and use a bitmap to eliminate from a calculation of absolute distances any plurality of values that are not present in all vectors.
    [15]
    15. System according to claim 11, characterized by the fact that the one or more processors are configured to interact with the computer-readable medium to perform operations that include:
    identify, for each respective Euclidean measurable biometric vector, a plurality of floating point values; and define a sliding scale of importance based on the number of vectors, a respective of the floating point value appears.
    [16]
    16. System according to claim 11, characterized by the fact that the neural network is configured with a variety of convolutional layers, together with a rectifier (ReLU) and cluster nodes.
    [17]
    17. System, according to claim 11, characterized by the fact that the neural network is configured to use clustering as a form of non-linear sampling,
    Petition 870200002745, of 07/01/2020, p. 13/17
    9/10 and in which one or more cluster nodes progressively reduce the spatial size of a Euclidean measurable characteristic vector represented to reduce the number of parameters and computation in the neural network.
    [18]
    18. System, according to claim 15, characterized by the fact that one or more processors are configured to interact with the computer-readable medium, in order to perform operations that include:
    calculating, for each of a plurality of stored Euclidean measurable characteristic vectors, a relative position difference between a medium face vector and the respective Euclidean measurable characteristic vector;
    quadrature of the relative position difference;
    add the values; and calculate the square root.
    [19]
    19. System, according to claim 11, characterized by the fact that the performance of the neural network is determined as a function of a cost function, in which a number of layers given as a spatial size of an output volume is computed as a function of an input volume size W, the size of the nucleus field of the K layer neurons, a step with which the layers are
    Petition 870200002745, of 07/01/2020, p. 14/17
    10/10 applied S and a zero fill amount P used on an edge.
    [20]
    20. System, according to claim 11, characterized by the fact that the neural network translates the initial biometric vector and the current biometric vector as a function of matrix multiplications for each respective layer and uses a Euclidean distance algorithm based on a Euclidean cost function.
    类似技术:
    公开号 | 公开日 | 专利标题
    BR112019023750A2|2020-05-26|SYSTEM AND METHOD FOR BIOMETRIC IDENTIFICATION
    Zhang et al.2017|Pic: Enable large-scale privacy preserving content-based image search on cloud
    US10764054B2|2020-09-01|Public/private key biometric authentication system
    JP6037366B2|2016-12-07|Method for authenticating user corresponding to encrypted data and system for authenticating user corresponding to biometric data
    US10009179B2|2018-06-26|Trusted platform module | protected device
    CN104951680B|2019-06-07|A kind of biological information processing method, store method and device
    Nagaraju et al.2015|Trusted framework for online banking in public cloud using multi-factor authentication and privacy protection gateway
    Zheng et al.2017|Privacy-preserving image denoising from external cloud databases
    US9916432B2|2018-03-13|Storing and retrieving cryptographic keys from biometric data
    US10740448B2|2020-08-11|Verification request authentication machine
    Wu et al.2016|High-dimension space projection-based biometric encryption for fingerprint with fuzzy minutia
    US20220069991A1|2022-03-03|System and Method for Securing Personal Information Via Biometric Public Key
    Yang et al.2019|A cancelable iris-and steganography-based user authentication system for the Internet of Things
    US10083194B2|2018-09-25|Process for obtaining candidate data from a remote storage server for comparison to a data to be identified
    US10853459B2|2020-12-01|Verification request authentication machine
    Karuppiah et al.2021|Secured storage and disease prediction of E-health data in cloud
    Bringer et al.2013|Practical identification with encrypted biometric data using oblivious ram
    Bhargav-Spantzel et al.2010|Biometrics-based identifiers for digital identity management
    Abduljabbar et al.2019|EEIRI: efficient encrypted image retrieval in IoT-cloud
    Tran et al.2017|Improving gait cryptosystem security using Gray code quantization and linear discriminant analysis
    Hu et al.2020|SensIR: Towards privacy-sensitive image retrieval in the cloud
    US10764055B1|2020-09-01|Cluster-based security for network devices
    KR20190118414A|2019-10-18|Method and apparatus for generating cryptographic key using biometric information
    Tran et al.2018|A Binarization Method for Extracting High Entropy String in Gait Biometric Cryptosystem
    Bradish et al.2021|CoviChain: A Blockchain Based COVID-19 Vaccination Passport
    同族专利:
    公开号 | 公开日
    AU2018266602A1|2019-12-12|
    CN110892693A|2020-03-17|
    WO2018208661A9|2019-05-09|
    CA3063126A1|2018-11-15|
    EP3635937A4|2021-01-20|
    EP3635937A1|2020-04-15|
    CO2019013817A2|2020-01-17|
    WO2018208661A1|2018-11-15|
    US20180330179A1|2018-11-15|
    JP2020520509A|2020-07-09|
    ZA201907621B|2021-04-28|
    US10255040B2|2019-04-09|
    KR20200007010A|2020-01-21|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US5161204A|1990-06-04|1992-11-03|Neuristics, Inc.|Apparatus for generating a feature matrix based on normalized out-class and in-class variation matrices|
    EP1759330B1|2004-06-09|2018-08-08|Koninklijke Philips Electronics N.V.|Biometric template similarity based on feature locations|
    KR101185525B1|2005-07-29|2012-09-24|텔레콤 이탈리아 소시에떼 퍼 아찌오니|Automatic biometric identification based on face recognition and support vector machines|
    US20080298642A1|2006-11-03|2008-12-04|Snowflake Technologies Corporation|Method and apparatus for extraction and matching of biometric detail|
    US8041956B1|2010-08-16|2011-10-18|Daon Holdings Limited|Method and system for biometric authentication|
    WO2012123727A1|2011-03-11|2012-09-20|Callsign, Inc|Personal identity control|
    US8942431B2|2012-06-24|2015-01-27|Neurologix Security Group Inc|Biometrics based methods and systems for user authentication|
    US8369595B1|2012-08-10|2013-02-05|EyeVerify LLC|Texture features for biometric authentication|
    US8811772B2|2012-08-21|2014-08-19|Tianzhi Yang|Mapping evaluating for spatial point sets|
    US9003196B2|2013-05-13|2015-04-07|Hoyos Labs Corp.|System and method for authorizing access to access-controlled environments|
    US9838388B2|2014-08-26|2017-12-05|Veridium Ip Limited|System and method for biometric protocol standards|
    IL231862A|2014-04-01|2015-04-30|Superfish Ltd|Neural network image representation|
    US20150356704A1|2014-06-08|2015-12-10|Yeda Research And Development Co. Ltd.|Good planar mappings and controlling singular values with semidefinite programming|
    CN104616032B|2015-01-30|2018-02-09|浙江工商大学|Multi-camera system target matching method based on depth convolutional neural networks|
    CN104766062A|2015-04-07|2015-07-08|广西大学|Face recognition system and register and recognition method based on lightweight class intelligent terminal|
    JP6416438B2|2015-09-11|2018-10-31|アイベリファイ インコーポレイテッド|Image and feature quality for ocular blood vessel and face recognition, image enhancement and feature extraction, and fusion of ocular blood vessels with facial and / or sub-facial regions for biometric systems|
    JP6504013B2|2015-10-13|2019-04-24|富士通株式会社|Cryptographic processing method, cryptographic processing device, and cryptographic processing program|
    US20170124263A1|2015-10-30|2017-05-04|Northrop Grumman Systems Corporation|Workflow and interface manager for a learning health system|
    US9846800B2|2015-11-16|2017-12-19|MorphoTrak, LLC|Fingerprint matching using virtual minutiae|
    CN105553980A|2015-12-18|2016-05-04|北京理工大学|Safety fingerprint identification system and method based on cloud computing|
    CN105956532B|2016-04-25|2019-05-21|大连理工大学|A kind of traffic scene classification method based on multiple dimensioned convolutional neural networks|
    JP6850817B2|2016-06-03|2021-03-31|マジック リープ， インコーポレイテッドＭａｇｉｃ Ｌｅａｐ，Ｉｎｃ．|Augmented reality identification verification|
    CN106227851B|2016-07-29|2019-10-01|汤一平|The image search method of depth of seam division search based on depth convolutional neural networks|
    JP6900664B2|2016-12-14|2021-07-07|富士通株式会社|Image processing equipment, image processing method and image processing program|US11163983B2|2012-09-07|2021-11-02|Stone Lock Global, Inc.|Methods and apparatus for aligning sampling points of facial profiles of users|
    US11017212B2|2012-09-07|2021-05-25|Stone Lock Global, Inc.|Methods and apparatus for biometric verification|
    US11017211B1|2012-09-07|2021-05-25|Stone Lock Global, Inc.|Methods and apparatus for biometric verification|
    US11163984B2|2012-09-07|2021-11-02|Stone Lock Global, Inc.|Methods and apparatus for constructing biometrical templates using facial profiles of users|
    US11017213B1|2012-09-07|2021-05-25|Stone Lock Global, Inc.|Methods and apparatus for biometric verification|
    US11017214B1|2012-09-07|2021-05-25|Stone Lock Global, Inc.|Methods and apparatus for biometric verification|
    US9838388B2|2014-08-26|2017-12-05|Veridium Ip Limited|System and method for biometric protocol standards|
    US10713535B2|2017-09-15|2020-07-14|NovuMind Limited|Methods and processes of encrypted deep learning services|
    WO2021055380A1|2019-09-17|2021-03-25|Private Identity Llc|Systems and methods for privacy-enabled biometric processing|
    US11138333B2|2018-03-07|2021-10-05|Private Identity Llc|Systems and methods for privacy-enabled biometric processing|
    US11265168B2|2018-03-07|2022-03-01|Private Identity Llc|Systems and methods for privacy-enabled biometric processing|
    US11210375B2|2018-03-07|2021-12-28|Private Identity Llc|Systems and methods for biometric processing with liveness|
    US11170084B2|2018-06-28|2021-11-09|Private Identity Llc|Biometric authentication|
    US11079911B2|2018-12-26|2021-08-03|Synaptics Incorporated|Enrollment-free offline device personalization|
    US11201745B2|2019-01-10|2021-12-14|International Business Machines Corporation|Method and system for privacy preserving biometric authentication|
    CN110048827B|2019-04-15|2021-05-14|电子科技大学|Class template attack method based on deep learning convolutional neural network|
    WO2022015948A1|2020-07-15|2022-01-20|Georgia Tech Research Corporation|Privacy-preserving fuzzy query system and method|
    US10938852B1|2020-08-14|2021-03-02|Private Identity Llc|Systems and methods for private authentication with helper networks|
    法律状态:
    2021-10-19| B350| Update of information on the portal [chapter 15.35 patent gazette]|
    2022-03-03| B08F| Application dismissed because of non-payment of annual fees [chapter 8.6 patent gazette]|Free format text: REFERENTE A 4A ANUIDADE. |
    优先权:
    申请号 | 申请日 | 专利标题
    US15/592,542|2017-05-11|
    US15/592,542|US10255040B2|2017-05-11|2017-05-11|System and method for biometric identification|
    PCT/US2018/031368|WO2018208661A1|2017-05-11|2018-05-07|System and method for biometric identification|
    [返回顶部]