CN114168918A

CN114168918A - Face information protection and bidirectional authentication system based on PUF

Info

Publication number: CN114168918A
Application number: CN202111428444.1A
Authority: CN
Inventors: 李冰; 马茜雅
Original assignee: Shenzhen Institute Of Southeast University
Current assignee: Shenzhen Institute Of Southeast University
Priority date: 2021-11-29
Filing date: 2021-11-29
Publication date: 2022-03-11

Abstract

The invention discloses a face information protection and bidirectional authentication system based on PUF, which comprises a revocable face template generation stage, a registration stage, a client verification stage and a server verification stage; the revocable face template generation stage comprises a feature extraction module and a random replacement module; the registration phase comprises an auxiliary information generation module; the client verification stage comprises an error correction code module and a Hash verification module; the server side verification stage comprises a template restoration module and a template matching module; the invention can revoke the combination of the biological template and the PUF, can realize the bidirectional identity authentication between the terminal equipment and the terminal user, and can resist the leakage of face information.

Description

Face information protection and bidirectional authentication system based on PUF

Technical Field

The invention relates to a face information protection and bidirectional authentication system based on PUF (physical unclonable function), which is used for realizing bidirectional identity verification between terminal equipment and a terminal user and belongs to the field of identity authentication based on biological characteristics.

Background

With the progress of society and the development of information technology, people pay more and more attention to the problem of information security, and the identity authentication of legal users is of great importance in various fields. In a conventional identity authentication system, physical objects such as keys, certificates, smart cards, etc. are used for identity authentication, and these entities are at risk of being lost or stolen. The cryptography technology provides powerful guarantee for information security, and the secret key is widely applied to various authentication systems: if the secret key is very short, the safety of the identity authentication mechanism is very weak, and guessing attack and brute force attack are easy to happen; if the key is long, it is difficult for the user to remember it accurately. Biometric identification technology is also increasingly widely applied to identity authentication systems in various fields, and is used as the physical identity of a user in the fields of internet of things, block chains, cloud computing and the like. In recent years, privacy and security of biometric data have received increasing attention. The biometric data is permanently associated with the identity of the user, and once stolen, sensitive information is leaked, and the biometric information of the user cannot be revoked, so that irreversible serious loss is caused. The method has the advantages that the revocable biological template is provided, original biological information is protected, the original biological information is combined with random numbers, and revocable and irreversibility are achieved through certain guard transformation. These random numbers still need to be stored in the physical device and there is also a risk of theft and loss.

Each person's biometric information is unique, and a PUF is a physical "fingerprint" that is unique to the device. Each transistor in an integrated circuit may produce a measurable output difference due to process variations in deep sub-micron manufacturing processes, which physical difference is unclonable and unpredictable to replicate. Only one response output exists at one excitation input, so that the PUF can assist the storage of secret data, and the protection data is not directly stored in a nonvolatile memory (NVM), so that the traditional physical attack is resisted, and the information leakage is avoided.

Disclosure of Invention

The invention aims to solve the technical problem of providing a face information protection and bidirectional authentication system based on PUF, which generates random number seeds through a PUF of user equipment to generate a revocable face biological template without storing random numbers to avoid the attack of an attacker on face information. Meanwhile, the uniqueness of the PUF as a physical fingerprint is utilized to complete the verification of the client to the server. In addition, the PUF at the server side is utilized to assist in protecting the storage of the revocable biological template of the user, so that the leakage of information is avoided, and the original biological information of the user is further protected.

In order to solve the technical problems, the face information protection and bidirectional authentication system based on the PUF comprises a revocable face template generation stage, an enrollment stage, a client verification stage and a server verification stage.

The revocable face template generation stage comprises a feature extraction module and a random replacement module, wherein the feature extraction module is used for extracting the features of the collected face image and extracting the features into 512-bit binary codes, in order to improve the performance of feature extraction in the aspects of ambient illumination change, face direction change, face shielding and the like, the feature extraction module extracts feature vectors from the face image by using a FaceNet-based depth CNN model, generates face vectors through L2 normalization and finally extracts the face vectors into 512-dimensional feature vectors. The random replacement module is used for generating a random number for performing replacement operation on the original biological template, and generating the revocable face template by taking the PUF response as a random number seed. In order to protect the random number of the revocable biometric template from being stolen, the random replacement module takes the PUF response as a generation seed for replacing the random number according to the CRPs of the PUF in the client, and only the PUF stimulus is stored in the client database. Generating a random number by taking the PUF response as a seed of a random number generator, and finishing the replacement operation of the original biological template to generate a revocable biological template; the user only keeps PUF excitation locally, so that the risk that random numbers are stolen is avoided, and an attacker cannot reverse the original biological template; at the same time, the revocable of the biological template can be completed by replacing CRP of PUF.

The registration stage comprises an auxiliary information generation module, and the auxiliary information generation module is used for generating auxiliary information which is required by the verification stage and can restore the revocable face template in the registration stage, so that the hiding of the face information storage stage is realized, and information leakage is resisted.

The client verification stage comprises an error correction code module and a hash verification module, wherein the error correction code module is used for removing noise in PUF response and realizing stable generation of 512-bit PUF response; in order to correct the influence on the PUF response caused by noise and the like, the error correction code module adopts a double-layer error correction technology combining a Hadamard code and a Reed-Solomon code to correct the influence on the PUF response caused by the noise, so that the generation of the revocable template is completed when the user verifies. The Hash verification module is used for verifying the authenticity of the server by the user and realizing the identity verification of the server by the user. The hash verification module adopts an MD5 message digest algorithm.

The server side verification stage comprises a template restoring module and a template matching module, wherein the template restoring module is used for restoring the revocable biological characteristics of the server side in the registration stage; and the template matching module is used for verifying the identity information of the user by the server and determining whether the identity information passes the verification or not by comparing the Hamming distances of the two groups of biological templates.

The face information protection and bidirectional authentication system based on the PUF comprises the following steps:

revocable face template generation stage: (1) the method comprises the steps that a camera collects a face image, the face image is extracted by adopting a depth CNN framework, and 512-dimensional binary face vectors are generated through L2 normalization; (2) the PUF of the embedded equipment of the user randomly generates a group of CRPs, the response of the CRPs is used as the seed of a random number to generate 512-bit random numbers, the sizes of the random numbers are randomly arranged in (1,512), and the binary face vector is subjected to replacement operation to generate a revocable face biological template T;

a registration stage: step A, sending a user ID, a revocable face biological template T and a hash value H (R) of a user PUF response to a server; b, randomly generating a group of CRPs by an embedded device PUF at a server end, carrying out exclusive or operation on the response of 512 bits and Tb after the revocable face biological template is coded by a Reed-Muller code to generate auxiliary information HelpData, reserving Challenge (S), and deleting Reverse (S); step C, packaging and storing the user ID, the auxiliary information HelpData, the service terminal challenge (S) and the hash value H (R) of the response of the user PUF in a memory; step D, after the registration is finished, the client only stores challenge (C) and the user ID;

a client verification stage: (i) a user sends a verification request to a server, and the sending content comprises a user ID and a random number N; (ii) after receiving the verification request, the server side searches a hash value H (R) corresponding to the user ID in the database, performs hash operation on the H (R) and the random number N to obtain HS, and sends the HS to the client side; (iii) the client PUF obtains response (C) after double error correction according to the locally stored challenge (C) as an incentive, calculates a hash value H (R) of the response, further calculates a hash value HC of H (R) and N, compares whether HC is the same as HS in the step (ii), and if the HC is the same as HS in the step (ii), the client verification is successful;

and a server side verification stage: (a) after the client successfully verifies, the client generates a revocable biological template T 'according to the newly acquired face image and response (C) as a replacement random number, and sends the revocable biological template T' to the server for verification request; (b) generating PUF response (S) (response) according to the challenge (S) corresponding to the user ID, carrying out exclusive OR on the response (S) (response) and the auxiliary information HelpData, and decoding and restoring the revocable biological template T when the registration is carried out through a Reed-Muller code; (c) the server side makes a Hamming distance between the received revocable biological template T' and the restored revocable biological template T, if the Hamming distance is smaller than the threshold tau, the server side is successfully verified, otherwise, the server side fails.

Only the auxiliary information HelpData is stored instead of directly storing the revocable biological template, so that the identity information of the user is protected, and the information leakage is resisted. Meanwhile, through the encoding and decoding of the error correcting code, the fluctuation of the PUF caused by the influence of noise and the like can be eliminated.

Compared with the prior art, the invention has the following beneficial technical effects:

(1) because the second step of the generation stage of the revocable face template is adopted, the response of the PUF of the user equipment is used as the seed generated by the random number, the random number does not need to be directly stored in the memory, the attack that an attacker steals the random number to restore the original face template is resisted, the revocable property of the template is simultaneously completed, and the random number can be replaced by replacing the response of the PUF, namely the random number seed.

(2) Because the second step of the registration stage, the third step of the client verification stage and the second step of the server verification stage all adopt an error correction code mechanism to obtain stable PUF response output, the accuracy of the biological characteristic identification technology is improved, and the failure of verification caused by the influence of noise and the like is avoided.

(3) Due to the three steps of the client verification stage, the identity verification of the client to the server is completed by responding the hash value and combining the hash value with the random number, and an attacker is prevented from simulating the server to repeatedly collect the personal template information of the user.

(4) Due to the three steps of the server authentication phase, the PUF response and the auxiliary information are used for restoring the user revocable biological template information in the registration phase. And the user template information is not directly stored, so that the leakage of the user information is avoided. Meanwhile, as the PUF is unique and not clonable, the PUF response is not stored, and an attacker cannot deduce the template information of the user according to the auxiliary information.

Drawings

The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

Fig. 1 is a detailed step diagram of a revocable face template generation stage and an enrollment stage of a PUF-based face information protection and mutual authentication system.

Fig. 2 is a diagram illustrating specific steps of a client verification stage and a server verification stage of the PUF-based face information protection and mutual authentication system.

Detailed Description

The invention relates to a face information protection and bidirectional authentication system based on PUF, which mainly comprises four parts of a revocable face template generation stage, an enrollment stage, a client verification stage and a server verification stage.

Fig. 1 is a diagram of specific steps in a revocable face template generation stage and an enrollment stage, as shown in fig. 1, a user device PUF is used as a seed of a random number generator, a feature vector obtained by face image feature extraction is subjected to a replacement operation according to a random number, and an exclusive or is performed on a generated revocable biometric template and a server PUF encoded by a Reed-Muller code to obtain auxiliary information. And finally, storing the user ID, the response hash value of the user equipment PUF and the auxiliary information HelpData into a nonvolatile memory of the server. The method comprises the following specific steps:

step A1: feature vectors are extracted from the face image using a FaceNet based depth CNN model and 512-dimensional binary face vectors are generated by L2 normalization.

Step A2: 512 bits are seeded by the user device PUF to generate a random number array with each bit value between (1,512), and this random array is permuted with the binary face vector to generate a revocable face biometric template, while the stimulus and response hash values for the response are computed and retained.

Step A3: and the response of the PUF at the server side is subjected to exclusive OR operation with the revocable face biological template in the step A2 after being coded by the Reed-Muller code, and HelpData is generated. And finally, storing the PUF stimulus, the hash value of the response of the user device PUF in the step A2, HelpData and the user ID into the NVM of the server.

The face information protection and bidirectional authentication system based on the PUF adopts the PUF response as the seed of the random number generator, so that the safety of the random number can be ensured, namely the random number cannot be stolen or changed. Meanwhile, the revocable biological template is irreversible and revocable according to the position permutation of the random number aiming at the original biological template.

Fig. 2 is a diagram of specific steps in a client authentication phase and a server authentication phase, where in the client authentication phase, a user device PUF generates a response according to an excitation, generates a stable PUF response through an error correction code, and sends a request for authentication of a user ID and a random number N to the server, the authenticity of the server is authenticated by comparing Hash (r) i N) transmitted by the server, and a biometric template to be authenticated is sent after authentication is successful, as shown in fig. 2. And in the verification stage of the server, the server PUF generates a response according to the excitation, the response is subjected to exclusive OR with the auxiliary information and then is subjected to error correction code decoding to recover the revocable biological template, the hamming distance is matched with the verification template sent by the user, and the verification is successful when the hamming distance is smaller than a threshold value. The method comprises the following specific steps:

step B1: and the user sends an authentication request to the server and simultaneously sends the user ID and the random number N to authenticate the authenticity of the server. The PUF of the user generates a Response according to the stored excitation, and the Response 'is obtained through double-layer error correction combining a Hadamard code and a Reed-Solomon code, and further the Hash value Hash (R)'.

Step B2: after receiving the user ID and the random number N, the server inquires a user PUF response hash value stored in the database, sends a verification formula to the client, and if the following formula is met, the verification is successful:

Hash(Hash(R)’||N)＝Hash(Hash(R)||N)

step B3: and B1, using the Response 'in the step B1 as a seed of a random number generator to generate a random number, and performing replacement operation on the random number and a feature vector acquired according to the new face image in the authentication process to generate the revocable biometric template T'.

Step B4: and after the client successfully verifies, the user sends the revocable biological template T' of the user to the server.

Step B5: the server PUF generates a response according to the stored stimulus, and restores the revocable biological template T when the server PUF is registered after being subjected to exclusive OR with the auxiliary information HelpData through Reed-Muller code decoding operation, and if the following formula is met, the verification is successful:

Hamming(T,T’)<τ

the face information protection and bidirectional authentication system based on the PUF adopts Hadamard codes, Reed-Solomon codes and Reed-Muller codes as error correcting codes. The Hadamard code and the Reed-Solomon code are used for response correction of the client PUF, a double code correction technology is adopted because high accuracy is required, and the PUF as a random number seed has no deviation, otherwise, the authentication of a real user fails. When the revocable biological template of the server side is recovered, the Reed-Muller code is adopted, the error correction capability of the system can be flexibly designed through the adjustment of R, K, and response errors caused by noise are solved.

The CRP, uniqueness and unclonability of the PUF are adopted in the face information protection and bidirectional authentication system based on the PUF, so that the random number when the template is generated and the biological characteristic template stored when the server side is registered are not directly stored in the memory, the attack of an attacker and the information leakage are avoided, and the bidirectional identity authentication is completed through the PUF.

The above embodiments do not limit the present invention in any way, and all technical solutions obtained by means of equivalent substitution or equivalent transformation fall within the protection scope of the present invention.

Claims

1. based on the PUF-based face information protection and two-way authentication system, it is characterized in that comprising revocable face template generation stage, registration stage, client verification stage, server verification stage;

The revocable face template generation stage includes a feature extraction module and a random replacement module; the registration stage includes an auxiliary information generation module; the client verification stage includes an error correction code module and a hash verification module; the server verification The stage includes template restoration module and template matching module;

The feature extraction module is used to extract the collected face image features, and the features are extracted as 512-bit binary codes; the random replacement module is used to generate a random number for performing a replacement operation on the original biological template, and responds with PUF As a random number seed, realize the generation of a revocable face template;

The auxiliary information generation module is used in the registration stage to generate auxiliary information required in the verification stage that can restore the revocable face template, so as to hide the face information in the storage stage and resist information leakage;

The error correction code module is used to complete the removal of noise in the PUF response and realize the generation of a stable 512-bit PUF response; the hash verification module is used for the user to verify the authenticity of the server and realize the user's identity verification to the server ;

The template restoration module is used for the server to restore the revocable biometrics in the registration stage; the template matching module is used for the server to verify the user identity information.

2. the facial information protection and bidirectional authentication system based on PUF according to claim 1, it is characterized in that: described feature extraction module uses the deep CNN model based on FaceNet to extract feature vector from facial image, and by L2 normalization. The face vector is generated by a normalization, and finally extracted as a 512-dimensional feature vector.

3. The face information protection and bidirectional authentication system based on PUF according to claim 1, it is characterized in that: described random replacement module according to the CRPs of PUF in the client, the PUF response is used as the generation seed that will replace random number, Only PUF incentives are stored in the client database.

4. the face information protection and bidirectional authentication system based on PUF according to claim 3, it is characterized in that: generate random number with PUF response as the seed of random number generator, complete the replacement operation to the original biological template to generate the random number. Revoked biological template; the user only retains the PUF incentive locally, avoiding the risk of random number being stolen, and the attacker cannot reverse the original biological template; at the same time, the revocability of the biological template can be completed by replacing the CRP of the PUF.

5. the facial information protection based on PUF according to claim 1 and the two-way authentication system, it is characterized in that: described error correction code module adopts the double-layer error correction technology that Hadamard code and Reed-Solomon code combine, corrects the cause of the error. The effect of noise on the PUF response to complete the generation of revocable templates during user authentication.

6 . The PUF-based face information protection and bidirectional authentication system according to claim 1 , wherein the hash verification module adopts an MD5 message digest algorithm. 7 .

7. the facial information protection based on PUF according to claim 1 and the bidirectional authentication system, it is characterized in that comprising the steps:

The revocable face template generation stage: (1) The camera collects the face image, uses the deep CNN architecture to extract the face image, and generates a 512-dimensional binary face vector through L2 normalization; (2) The user's embedded device PUF Randomly generate a set of CRPs, and use the responses as the seeds of random numbers to generate 512-bit random numbers. The random numbers are randomly arranged in (1,512), and perform a replacement operation on the binary face vector to generate a revocable face. biological template T;

Registration stage: Step A, send the user ID, the revocable face biological template T, and the hash value H(R) of the user's PUF response to the server; Step B, the embedded device PUF on the server randomly generates a set of CRPs, The 512-bit response and the revocable face biological template are XORed with the Tb encoded by the Reed-Muller code to generate auxiliary information HelpData, retain Challenge(S), and delete Reponse(S); Step C, the user ID , auxiliary information HelpData, server Challenge (S), the hash value H (R) of user PUF response are packaged and stored in the memory; Step D, after the registration is completed, the client only saves Challenge (C), user ID;

Client verification stage: (i) the user sends a verification request to the server, and the sent content includes the user ID and a random number N; (ii) after receiving the verification request, the server searches the database for the hash value corresponding to the user ID H(R), hash H(R) and random number N together to get HS, and send the HS back to the client; (iii) The client PUF uses the locally saved Challenge(C) as an incentive, and sends the response through double After error correction, Response(C)' is obtained, the hash value H(R)' of the response is calculated, and the hash value HC of H(R)' and N is further calculated, and HC and HS in step (ii) are compared whether they are the same , if they are identical, the client authentication is successful;

Server-side verification stage: (a) After the client-side verification is successful, the client-side generates a revocable biological template T' according to the newly collected face image and Response(C)' as a replacement random number, and sends it to the server-side for verification; ( b) Generate the PUF response Response(S)' according to the Challenge(S) corresponding to the user ID, XOR the Response(S)' with the auxiliary information HelpData, and restore the revocable biological template T during registration through Reed-Muller code decoding; ( c) The server performs the Hamming distance between the received revocable biological template T' and the restored revocable biological template T. If the distance is less than the threshold τ, the verification of the server is successful, otherwise it fails.