METHOD AND SYSTEM FOR PRIVACY-PROTECTING BIOMETRIC AUTHENTICATION

A system using homomorphic encryption and zero-knowledge proofs securely authenticates encrypted biometric data with a neural network classifier, addressing detection performance and security issues in existing methods, and achieving balanced performance and security.

DE112019006622B4Active Publication Date: 2024-11-14INTERNATIONAL BUSINESS MACHINE CORPORATION
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
DE112019006622
Authority / Receiving Office
DE · DE
Patent Type
Patents
Current Assignee / Owner
Priority Date
2019-01-10
Filing Date
2019-12-13
Publication Date
2024-11-14
Estimated Expiration
2039-12-13

AI Technical Summary

Technical Problem

Current methods for encrypting biometric information for authentication suffer from unacceptable reductions in detection performance and lack provable security, making biometric databases vulnerable to attacks.

Method used

Implement a system using homomorphic encryption and zero-knowledge proofs to securely store and authenticate encrypted biometric data with a neural network classifier, where the neural network weights are trained with both public and private biometric information, and authentication is performed using encrypted match scores verified by a client device with a private key.

Benefits of technology

Achieves authentication with undiminished recognition performance while ensuring security comparable to cryptographic standards, balancing memory and speed requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

Method (200) for biometric authentication, implemented in a computer comprising a processor (402A, ... 402N), a memory (408) accessible by the processor and computer program instructions stored in the memory (408) and executable by the processor (402A, ... 402N), wherein the method comprises: Storing a large number of encrypted trained weights (420) of a neural network classifier (108), wherein the weights were trained using biometric information representing at least one biometric feature of a person. Receiving encrypted biometric information obtained by scanning at least one biometric feature of the person, and encrypting the scanned biometric feature; Determining a match score using the encrypted trained weights to operate the trained neural network classifier (108), wherein the match score indicates a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein determining the match score involves: Determining an encrypted match score using the neural network classifier (108); Transmitting the encrypted match score to a client device; Receiving (214) an encrypted match score from the client device; and Determine that the client device correctly decrypted the match score (304); and Authenticating the person when the probability that the received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold (308); and wherein determining the match score further includes: Multiply the encrypted match score by a first encrypted secret integer; Encrypting (302) a multitude of additional secret integers; Transmit (120) the multiplied encrypted match score and at least some of the multiple encrypted additional secret integers to a client device; Receiving a decrypted multiplied match score and decrypted additional secret integers from the client device; Verifying the correctness of the decrypted additional secret integers; and Divide the unencrypted match score by the first secret integer to obtain the match score.
Need to check novelty before this filing date? Find Prior Art

Description

BACKGROUND

[0001] The present invention relates to methods that provide encrypted biometric information which can be stored and used for authentication with undiminished recognition performance.

[0002] Unlike deterministic passwords, biometric characteristics constantly change due to factors such as variations in the scanning process (scanning noise, shadows, positional changes, etc.) or natural causes (injury, beard growth, age, etc.). Therefore, unlike passwords, biometric information will not be exactly the same on every scan and cannot be treated like passwords, e.g., stored, distorted, or protected using a hashing algorithm. Currently, there is no practical method for encrypting biometric information and preserving its value for authentication. One reason for this lack of availability is, for example, the unacceptable reduction in recognition performance combined with unprovable security claims. Consequently, biometric information is currently stored in various databases that are vulnerable to attacks.

[0003] Several documents have already been published. Document US 2011 / 0 285 504 A1 describes a method and device for biometric identity verification. This involves a registration process in which reference data for a biometric attribute of a reference user is recorded to train a neural network. This trained neural network is then used to identify a person at an access point. Document US 2014 / 0 185 794 A1 describes an encryption device and a corresponding method. The encryption method used is homomorphic encryption. Finally, document US 2018 / 0 330 179 A1 describes a system and method for biometric identification. This involves comparing encrypted biometric input data with at least one set of stored biometric data.However, these known methods have at least some gaps in the entire encryption chain.

[0004] Accordingly, there is a need for techniques that provide encrypted biometric information that can be stored and used for authentication with undiminished recognition performance. SUMMARY

[0005] This problem is solved by the subject matter of the independent claims. Further embodiments are specified by the dependent claims. Embodiments of the present systems and methods can provide encrypted biometric information that can be stored and used for authentication with undiminished recognition performance. Embodiments can offer advantages over current techniques. For example, embodiments can offer security claims comparable to current cryptographic solutions, such as symmetric and asymmetric methods, since embodiments can incorporate known and accepted cryptographic modules. Furthermore, the reduction in recognition performance rates can be described as a trade-off between storage and speed requirements; and for industrially acceptable performance requirements, the embodiments can achieve both.

[0006] For example, in one embodiment, a biometric authentication method implemented in a computer comprising a processor, processor-accessible memory, and computer program instructions stored in the memory and executable by the processor, may include storing a plurality of encrypted, trained weights of a neural network classifier. These weights may be trained using biometric information representing at least one biometric characteristic of a person.The procedure may further include receiving encrypted biometric information obtained by recording at least one biometric characteristic of the person and encrypting the recorded biometric information, obtaining a match score using the encrypted trained neural network classifier, where the match score indicates a probability that the received encrypted biometric information matches the stored encrypted biometric information, and authenticating the person if the probability that the received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

[0007] In embodiments, the trained weights can be encrypted using homomorphic encryption, a multi-sided computation, or a combination of both. Retrieving the match score can involve: obtaining an encrypted match score using the encrypted, trained neural network classifier; transmitting the encrypted match score to a client device; receiving an unencrypted match score from the client device; determining that the client correctly decrypted the match score using a zero-knowledge proof; and comparing the unencrypted match score to the threshold.Obtaining the match score may involve: obtaining an encrypted match score using the encrypted trained neural network classifier, multiplying the encrypted match score by an initial encrypted secret integer, encrypting a multitude of additional secret integers, transmitting the multiplied encrypted match score and at least some of the multitude of encrypted additional secret integers to a client device, receiving a decrypted multiplied match score and decrypted additional secret integers from the client device, verifying the correctness of the decrypted additional secret integers, and dividing the unencrypted match score by the initial encrypted secret integer to obtain the match score.The multiplied encrypted match score and the encrypted additional secret integers can be transmitted to the client device in a secret random order. The encrypted biometric information can be received by a client device that is communicatively connected to a biometric data capture device. Decryption can be performed on the client device using a private key. Some layers of the neural network classifier can be trained on publicly available, non-private biometric information, and some layers of the neural network classifier can be retrained on the individual's private biometric information.

[0008] In one embodiment, a biometric authentication system may comprise a processor, memory accessible to the processor, and computer program instructions stored in the memory and executable by the processor to perform the following: storing multiple encrypted trained weights of a neural network classifier, wherein the weights were trained using biometric information representing at least one biometric feature of a person; receiving encrypted biometric information obtained by capturing at least one biometric feature of the person and encrypting the captured biometric feature; and obtaining a match score using the encrypted trained neural network classifier, wherein the match score represents a probability.that the received encrypted biometric information matches the stored encrypted biometric information, and authenticate the person if the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold.

[0009] In one embodiment, a computer program product for biometric authentication may have a non-transitory, computer-readable memory containing program instructions, wherein the program instructions are executable by a computer to cause the computer to perform a method comprising: storing multiple encrypted trained weights of a neural network classifier, wherein the weights were trained using biometric information representing at least one biometric feature of a person; receiving encrypted biometric information obtained by capturing at least one biometric feature of the person and encrypting the captured biometric feature; and obtaining a match score using the encrypted trained neural network classifier, wherein the match score represents a probability.that the received encrypted biometric information matches the stored encrypted biometric information, and authenticate the person if the probability that received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The details of the present invention, both in terms of its structure and operation, can best be understood by reference to the accompanying drawings, in which the same reference numerals and designations refer to the same elements. Fig. Figure 1 illustrates an exemplary system in which the embodiments of the present systems and methods can be implemented. Fig. Figure 2 is an exemplary flowchart of a process that can implement embodiments of the present methods and that can be implemented in embodiments of the present systems. Fig. Figure 3 is an exemplary flowchart of a process that can implement embodiments of the present methods and that can be implemented in embodiments of the present systems. Fig. Figure 4 is an exemplary block diagram of a computer system / computing device in which processes involved in the embodiments described here can be implemented. DETAILED DESCRIPTION

[0011] Embodiments of the present systems and methods can provide encrypted biometric information that can be stored and used for authentication with undiminished recognition performance. Embodiments can offer advantages over current techniques. For example, embodiments can provide security claims comparable to current cryptographic solutions, such as symmetric and asymmetric methods, since embodiments can incorporate known and accepted cryptographic modules. Furthermore, the reduction in recognition performance rates can be described as a trade-off between storage and speed requirements, and for industrially acceptable performance requirements, the embodiments can achieve both.

[0012] Leveraging advances in homomorphic encryption, implementations can include a system that stores an encrypted, user-specific neural classification network for each individual. Biometric data sent to the server can be encrypted and remain encrypted throughout the authentication process. Upon completion of the authentication process, the server can possess an encrypted match score. This score can only be decrypted by the user who holds the private key, which can then be transmitted to the server via a special protocol. The server can then use this key to either authenticate or reject the user.

[0013] An exemplary system 100, in which embodiments of the present systems and methods can be implemented, is described in Fig. Figure 1 illustrates this. In this example, the system 100 can include a server 102, a client device 104, and a device 106 for capturing biometric information. The server 102 can perform functions such as logging in and authenticating users of the system and can contain a variety of encrypted neural networks 108. The client device 104 can be any computer capable of running software programs and can include general-purpose computing devices such as a personal computer, laptop, smartphone, tablet computer, etc., as well as specialized computing devices such as embedded processors, systems on a chip, etc., which may be contained in standard or proprietary devices such as input devices, kiosks, ATMs, etc. The client device 104 can contain and run one or more client applications 110.The device 106 for capturing biometric information can be any device capable of capturing biological, physiological, and / or physical biometric information, such as a fingerprint, retinal scan, hand veins, facial recognition, DNA, palm print, hand geometry, iris recognition, retinal and odor / scent, etc. The server 102 can be communicatively connected to the client device 104 and the device 106 for capturing biometric information. The client device 104 can be communicatively and / or physically connected to the device 106 for capturing biometric information.

[0014] An exemplary flowchart of process 200 for the operation of system 100 is in Fig. 2 shown. It is best used in conjunction with Fig. 1. In this example, the process 200 can have two phases: the registration phase 202 and the verification phase 204. The registration phase 202 can begin at 206, in which the device 106 captures one or more biometric features of a person and can convert the capture into biometric information that is transmitted to the client device 104, 112. On the client device 104, the client application 110 can train a neural network classifier 208 to identify the person's biometric features using the biometric information.In some implementations, the neural network classifier, or a subset of its layers, can be trained with publicly available, non-private biometric information, while other layers can be retrained with a specific individual's private biometric information. To reduce memory and speed requirements, for example, the entire neural network model (NN model) can be trained using publicly available biometric data (which is not subject to privacy restrictions). The weights of, say, the first few layers can be fixed or stored, and then, during registration, the remaining layers can be retrained with private, individual-specific biometric data. This way, only the last few layers, for instance, need to be encrypted.During verification, the biometric features can first be fed into the unencrypted neural network layers, then encrypted and sent to the encrypted model. The trained weights can be encrypted, e.g., with homomorphic encryption, and transmitted to the server, which can store the encrypted neural network for that person. Homomorphic encryption allows computation to be performed on encrypted data, so that the results of the computation on the encrypted data are the same upon decryption as if the computation had been performed on the unencrypted or plaintext data.

[0015] The verification phase 204 can begin at 212, in which the biometric features are re-captured or recorded, and the biometric information can be transmitted to the client device 104, 116. The client device 104 can process the biometric information, encrypt it, and transmit the encrypted biometric information to the server 102, 118. At 214, the server 102 can use the encrypted weights and the received biometric information features to employ the neural network and obtain an encrypted match score. The server 102 now needs to know whether this match score exceeds a threshold or not.Since only client device 104 can decrypt the match score, a process 216 can be used in which client device 104 decrypts the match score using a private key and transmits it to the server in a trusted and secure manner. In embodiments, the encrypted match score can be decrypted by the user who possesses a private key using a zero-knowledge proof. Likewise, it can be verified that the user, who may not be trusted, correctly decrypted the match score using a zero-knowledge proof. Examples of such zero-knowledge proofs may include, but are not limited to, performing a multiplication without further addition, other common functions, f(secret_numberi), techniques such as vickery auctions, etc.A Vickery or highest-bidder auction is "zero-knowledge". For example, the second-highest bidder can be considered as the threshold, with subsequent bids serving as proof of zero knowledge, where the given score represents the highest bidder.

[0016] An example flowchart of process 216 is in Fig. 3 shown. It is best used in conjunction with Fig. 1 can be seen. Process 216 can begin at 302, where server 102 can multiply the encrypted match score by an encrypted secret integer R1. Server 102 can also multiply N-1 other secret integers R2, ..., R NEncrypt. Server 102 can transmit the multiplied values, including the multiplied encrypted match score and the multiplied (encrypted) numbers, to client device 104. The multiplied values ​​can be transmitted in a secret random order. Client device 104 can decrypt the multiplied values. Client device 104 can transmit the decrypted value back to server 102, which verifies the correctness of R2, ..., R. Ncan verify. For the remaining value, this value can be divided by R1 to obtain the decrypted match score. Server 102 can then compare the decrypted match score to a threshold to determine if there is a match. If there is a match, the person can be authenticated. Client device 104 can decrypt the multiplied values. Client device 104 can transmit the decrypted value back to server 102, 122, which verifies the correctness of R2, ..., R N This can be verified. The remaining value can be divided by R1 to obtain the decrypted match score. Server 102 can then compare the decrypted match score to a threshold to determine if there is a match. If there is a match, the person can be authenticated.

[0017] An exemplary block diagram of a computer system / computer device 402, in which the processes of the embodiments described here may be implemented, is shown in Fig. Figure 4 illustrates the following. The computer system / computing device 402 can be implemented with one or more programmed general-purpose computing systems, such as embedded processors, systems-on-a-chip, personal computers, workstations, server systems, and minicomputers or mainframes, mobile devices such as smartphones or tablets, or in distributed, networked computing environments. The computer system / computing device 402 can include one or more processors (CPUs) 402A-402N, an input / output circuit 404, a network adapter 406, and a memory 408. The CPUs 402A-402N execute program instructions to perform the functions of the communication systems and procedures presented. Typically, the CPUs 402A-402N are one or more microprocessors, such as an Intel Core® processor or an Arm® processor. Fig. Figure 4 shows an embodiment in which the computer system / computing device 402 is implemented as a single multiprocessor computer system / computing device in which several processors 402A-402N share system resources, such as the memory 408, the input / output circuit 404, and the network adapter 406. However, the present communication systems and methods also include embodiments in which the computer system / computing device 402 is implemented as a plurality of network computer systems, which may be single-processor computer systems / computing devices, multiprocessor computer systems / computing devices, or a mixture thereof.

[0018] The input / output circuit 404 provides the ability to input data into and output data from the computer system / computing device 402. For example, the input / output circuit can include input devices such as keyboards, mice, touchpads, trackballs, scanners, analog-to-digital converters, etc.; output devices such as video adapters, monitors, printers, biometric information capture devices, etc.; and input / output devices such as modems, etc. The network adapter 406 connects the device 400 to a network 410. The network 410 can be any public or proprietary LAN or WAN, including, but not limited to, the Internet.

[0019] Memory 408 stores program instructions executed by CPU 402 and data used and processed by CPU 402 to perform the functions of computer system / computer device 402. Memory 408 can include, for example, electronic devices such as RAM (Random-Access Memory), ROM (Read-Only Memory), PROM (Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, etc., as well as electromechanical devices such as magnetic disk drives, tape drives, optical disk drives, etc., which have an Integrated Drive Electronics (IDE) interface or a variation or improvement thereof, such as Enhanced IDE (EIDE) or Ultra-Direct-Memory-Access (UDMA), or a Small Computer System Interface (SCSI)-based interface or a variation or improvement thereof, such as Fast-SCSI, Wide-SCSI, Fast- and Wide-SCSI, etc., or Serial Advanced Technology Attachment (SATA) or a variation or improvement thereof, or a Fiber Channel Abitrated Loop (FC-AL) interface.

[0020] The contents of memory 408 can vary depending on the function for which the computer system / calculating device 402 is programmed. In the Fig. The example shown in Figure 4 illustrates exemplary memory contents representing routines and data for embodiments of the processes described above. However, a person skilled in the art would recognize that these routines, together with the memory contents relating to these routines, need not be contained within a single system or device, but can be distributed across a multitude of systems or devices based on known technical considerations. The communication systems and methods presented here can incorporate any and all such arrangements.

[0021] In the Fig. In the example shown, memory 408 is depicted for brevity as containing memory contents for a server 412 and memory contents for a client device 414, but typically the computer system / computing device 400 contains only one such memory content. In this example, the server 412 can contain login routines 416, verification routines 418, and encrypted neural network data 420. Similarly, in this example, the client device 414 can contain login routines 422, verification routines 424, and verification data 426. The login routines 416 can contain software routines for performing server login processes, as described above. The verification routines 418 can contain software routines for performing server verification processes, as described above.Encrypted neural network data 420 can contain encrypted data representing trained neural networks, as described above. Login routines 422 can include software routines for performing client device login processes, as described above. Verification routines 424 can contain software routines for performing client device verification processes, as described above. Verification data 426 can contain encrypted and decrypted data used by the client device during the verification process, as described above. The operating system 428 can provide all system functionality.

[0022] As in Fig.As shown in Figure 4, the communication systems and methods presented here can be implemented on a system or systems that provide multiprocessor, multitasking, multiprocess, and / or multithreaded computing, as well as on systems that only provide single-processor, single-threaded computing. Multiprocessor computing involves performing computations using more than one processor. Multitasking computing involves performing computations using more than one operating system task. A task is an operating system concept that refers to the combination of an executed program and the control information used by the operating system. Each time a program is executed, the operating system creates a new task for that program. The task acts like an envelope for the program, identifying the program with a task number and attaching other control information to it.Many operating systems, including Linux, UNIX®, OS / 2®, and Windows®, are capable of executing many tasks simultaneously and are known as multitasking operating systems. A task acts like an envelope for a program, identifying it with a task number and attaching other control information. Multitasking is the ability of an operating system to run more than one executable file at the same time. Each executable file runs in its own address space, meaning that the executables have no way of sharing memory. This has advantages, as it makes it impossible for one program to interfere with the execution of any other program running on the system. However, the programs have no way to exchange information except through the operating system (or by reading files stored in the file system).Multi-process computing is similar to multi-tasking computing, as the terms task and process are often used interchangeably, although some operating systems differentiate between the two.

[0023] The present invention can be a system, a method, and / or a computer program product at any possible level of technical integration. The computer program product can include a computer-readable storage medium (or media) containing computer-readable program instructions that cause a processor to execute aspects of the present invention. The computer-readable storage medium can be an actual device capable of managing and storing instructions for use by an instruction execution device.

[0024] The computer-readable storage medium may be, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of computer-readable storage media includes: a portable computer disk, a hard disk, working memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static read-only memory (SRAM), a portable compact disc read-only storage device (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically coded device, such as punched cards or raised structures in a groove on which instructions are stored, and any suitable combination of the foregoing.A computer-readable storage medium, such as the one used here, is not limited to transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses traveling through a fiber optic cable), or electrical signals transmitted through a cable.

[0025] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium or an external computer or storage device to appropriate computing / processing equipment via a network, such as the internet, a local area network, a wide area network, and / or a wireless network. The network may include copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage on a computer-readable storage medium within the respective computing / processing device.

[0026] Computer-readable program instructions for performing operations of the present invention can be instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state data, integrated circuit configuration data, or either source code or object code written in any combination of one or more programming languages, including an object-oriented programming language such as Smalltalk, C++, or the like, and procedural programming languages ​​such as the "C" programming language or similar. The computer-readable program instructions can be executed entirely on the user's computer, partially on the user's computer as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server.In the latter scenario, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be established with an external computer (e.g., via the internet using an internet service provider). In some embodiments, electronic circuits, such as programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), can execute the computer-readable program instructions by using state information from the computer-readable program instructions to personalize the electronic circuits to perform aspects of the present invention.

[0027] Aspects of the present invention are described here with reference to flowchart diagrams and / or block diagrams of processes, devices (systems), and computer program products according to embodiments of the invention. It is understood that each block of the flowchart diagrams and / or block diagrams, as well as combinations of blocks in the flowchart diagrams and / or block diagrams, can be implemented by computer-readable program instructions.

[0028] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or any other programmable data processing device to create a machine such that the instructions executed through the processor of the computer or other programmable data processing device create means to implement the functions / actions specified in the flowchart and / or block diagram.These computer-readable program instructions may also be stored in a computer-readable storage medium that can instruct a computer, a programmable data processing device and / or other devices to function in a certain manner, such that the computer-readable storage medium in which instructions are stored has a manufacturing item containing instructions that implement aspects of the function / action specified in the flowchart and / or block diagram block or blocks.

[0029] The computer-readable program instructions can also be loaded onto a computer, other programmable data processing device, or other device to initiate a series of operational steps that are executed on the computer, other programmable device, or other device to create a computer-implemented process, such that the instructions executed on the computer, other programmable device, or other device implement the functions / actions specified in the flowchart and / or block diagram block or blocks.

[0030] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this respect, each block in the flowchart or block diagrams can represent a module, segment, or part of instructions that includes one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions specified in the blocks may occur in a different order than that shown in the figures. For example, two blocks shown consecutively may actually be executed essentially simultaneously, or the blocks may sometimes be executed in reverse order, depending on the specific functionality.It is also noted that each block of the block diagrams and / or flowchart representation, as well as combinations of blocks in the block diagrams and / or flowchart representation, can be implemented by special hardware-based systems that perform the specified functions or actions, or execute combinations of special hardware and computer instructions.

[0031] Although specific embodiments of the present invention have been described, the person skilled in the art will understand that there are other embodiments that are equivalent to those described. Accordingly, it should be understood that the invention is not limited by the specific embodiments described, but only by the scope of the appended claims.

Claims

[1] A method (200) for biometric authentication implemented in a computer having a processor (402A, ... 402N), a memory (408) accessible by the processor, and computer program instructions stored in the memory (408) and executable by the processor (402A, ... 402N), the method comprising: Storing a plurality of encrypted trained weights (420) of a neural network classifier (108), wherein the weights were trained using biometric information representing at least one biometric characteristic of a person Receiving encrypted biometric information obtained by scanning at least one biometric characteristic of the person and encrypting the scanned biometric characteristic; Determining a match score using the encrypted trained weights to operate the trained neural network classifier (108), the match score indicating a probability that the received encrypted biometric information matches the stored encrypted biometric information, wherein determining the match score comprises: Determining an encrypted agreement score using the neural network classifier (108); Transmitting the encrypted match score to a client device; Receiving (214) an encrypted match score from the client device; and Determining that the client device has correctly decoded the match score (304); and Authenticating the person if the probability that the received encrypted biometric information matches the stored encrypted biometric information exceeds a threshold (308); and wherein determining the match score further comprises: Multiplying the encrypted match score by a first encrypted secret integer; Encrypting (302) a plurality of additional secret integers; transmitting (120) the multiplied encrypted match score and at least some of the plurality of encrypted additional secret integers to a client device; Receiving a decrypted multiplied match score and decrypted additional secret integers from the client device; Verifying the correctness of the decrypted additional secret integers; and Divide the unencrypted match score by the first secret integer to get the match score. [2] The method of claim 1, wherein the trained weights are encrypted using homomorphic encryption, multi-page computation, or a combination of both. [3] The method of claim 2, wherein determining the match score comprises: Determine that the client has correctly decrypted the match score using a zero-knowledge proof; and Comparing the unencrypted match score with the threshold. [4] The method of claim 1, wherein the multiplied encrypted match score, the encrypted additional secret integers are transmitted to the client device (104) in a secret random order. [5] The method of claim 1, wherein the encrypted biometric information is received from a client device (104) communicatively coupled to a biometric information capture device. [6] The method of claim 5, wherein the decryption (304) is performed on the client device (104) using a private key. [7] The method of claim 1, wherein some of the layers of the neural network classifier (108) were trained using publicly available non-private biometric information and some layers of the neural network classifier (108) were retrained using private biometric information of the individual. [8] A system for biometric authentication, the system comprising a processor (402A, ... 402N), a memory (408) accessible by the processor (402A, ... 402N), and computer program instructions stored in the memory (408) and executable by the processor (402A, ... 402N) to perform the method (200) according to any one of claims 1 to 7. [9] A computer program product for biometric authentication, the computer program product comprising a non-transitory computer-readable memory having program instructions embodied therein, the program instructions being executable by a computer to enable the computer to carry out a method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Biometric identity verification

    US20110285504A1

  • Encryption processing apparatus and method

    US20140185794A1

  • System and method for biometric identification

    US20180330179A1