JP2011253378A - Authentication device and authentication method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To propose an authentication device and an authentication method capable of preventing that fingerprint information on each individual is specified from registered information.SOLUTION: A fingerprint image 20 is sectioned into a plurality of areas ER, and a plurality of pieces of fingerprint identification information by area are generated, which shows angles to be formed by a ridge 21 with feature parts of fingerprints in the fingerprint image 20 and a reference line y for every area ER, respectively. A common template 30 at an angle level where an endpoint angle θ1 and a branching point angle θ2 of the pieces of fingerprint identification information by area are included is selected for every piece of fingerprint identification information by area from among a plurality of common templates represented for every area ER by unit of a predetermined angle range, and a registered ID pattern X by converting each piece of fingerprint identification information by area into corresponding common templates 30 is generated. Thus, collation processing is performed while altering the fingerprint identification information by area which shows features of the fingerprints of the user, it also becomes difficult to presume the fingerprint identification information by area from the registered ID pattern X, and specification of the fingerprint information on each individual is prevented.

Description

  The present invention relates to an authentication device and an authentication method, and is suitable for application to an authentication device that performs personal authentication of whether or not the user is a legitimate user using a fingerprint image, for example.

  In recent years, biometric authentication has been known that uses personal biometric information such as fingerprints, irises, blood vessel patterns, etc. as authentication information, which significantly improves security compared to conventional personal authentication using passwords. Yes.

  In such biometric authentication, for example, when using a fingerprint, a feature amount is extracted from the user's fingerprint in advance and registered as registration information in a database. When performing personal authentication, the authentication information is generated by extracting features from the fingerprint of the user who is going to authenticate, the authentication information is compared with the registered information, and the similarity between the two is verified. If a preset condition is satisfied, the user is authenticated as a registered valid user (see, for example, Patent Document 1).

JP-A-10-91769

  However, in such biometric authentication, since the fingerprint information unique to the individual is extracted and registered in the database as registration information, if such registration information is leaked, the personal information is registered from the registration information. There is a problem that fingerprint information that is unique and invariant throughout life may be identified.

  Therefore, the present invention has been made in view of the above points, and an object thereof is to propose an authentication apparatus and an authentication method that can prevent each individual's fingerprint information from being specified from registration information.

  In order to solve this problem, claim 1 of the present invention divides a fingerprint image into a plurality of areas, and a plurality of angles each showing an angle formed by a feature portion of the fingerprint in the fingerprint image and a reference line. A generating unit that generates fingerprint identification information for each area, and a plurality of common templates that represent all the angles assumed by the feature portion of the fingerprint and the reference line for each area in a predetermined angle range unit are stored. A common template of the angle range including the angle of the fingerprint identification information for each area is selected for each of the fingerprint identification information for each area from among the plurality of common templates. Conversion means for generating pattern information obtained by converting fingerprint identification information into the corresponding common template.

  In addition, according to a second aspect of the present invention, the converting means calculates the average value of the angle range of the common template for each area after converting the fingerprint identification information for each area into the common template, and calculates the average value. The pattern information represented by the above is generated.

  According to a third aspect of the present invention, the conversion unit reads an output bit string associated in advance according to the average value, and generates the pattern information represented by the output bit string. It is.

  According to a fourth aspect of the present invention, when the pattern information is stored in the storage means as registration information, an error correction coding process is performed on the pattern information.

  According to a fifth aspect of the present invention, a fingerprint image is divided into a plurality of areas, and a plurality of fingerprint identifications for each area each indicating an angle formed by a feature portion of the fingerprint in the fingerprint image and a reference line. A generation step of generating information by a generation unit, and a plurality of common templates in which all the angles assumed by the feature portion of the fingerprint and the reference line are expressed for each area in a predetermined angle range unit. From the above, the common template of the angle range including the angle of the fingerprint identification information by area is selected for each fingerprint identification information by area by the conversion means, and the common fingerprint identification information corresponding to each area is respectively corresponding to the common fingerprint identification information by area. And a conversion step for generating pattern information converted into a template.

  According to the first and fifth aspects of the present invention, since the fingerprint information indicating the characteristics of the fingerprint is altered, it is difficult to estimate the fingerprint identification information for each area from the pattern information. It is possible to prevent the fingerprint information from being specified.

It is the schematic where it uses for the rough description of the authentication method of this invention. It is the schematic which shows the structure of the authentication system of this invention. It is the schematic shown about the structure of the arrange | positioned area. It is the schematic where it uses for description of the end point and branch point of a fingerprint image, an end point direction and a branch point direction, an end point angle, and a branch point angle. It is the schematic where it uses for description of a collation threshold value. It is the schematic which shows the structure of common template information. It is the schematic where it uses for description of template angle level information. It is the schematic where it uses for description of an area | region determination table. It is the schematic which shows the structure of an output bit conversion table. It is the schematic which shows a series of process steps after receiving a fingerprint image until it produces | generates a registration ID pattern. It is the schematic where it uses for description of the evaluation type | formula which is the conditions of confidential information restoration. It is the table | surface which summarized the experiment specifications of the evaluation experiment. It is the graph which showed the relationship between the number of secret information elements, and a restoration rate. 5 is a graph showing a relationship between a false match rate (FMR) and a non-false match rate (FNMR).

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

(1) Outline of the Present Invention FIG. 1 shows an outline of an authentication method according to the present invention, which is divided into a registration process and a verification process. In this authentication method, for example, fingerprint information representing the characteristics of the user's fingerprint is used, and the fingerprint information of the user is acquired at the time of collation processing, and is registered based on the information generated from the fingerprint information. It can be authenticated whether or not the user is a legitimate user.

  Here, the fingerprint information is obtained by imaging a user's finger and dividing a fingerprint image obtained by predetermined image processing into a predetermined number of areas, and determining whether there are fingerprint end points and branch points (maneuvers) for each area. Judgment and collection of fingerprint identification information (described later) for each area in which these end points, branch points, and angles thereof (described later) are represented by a predetermined number of bits (binary system of “0” and “1”). It is. In practice, in this authentication method, common template information, which is a feature of the present invention, is generated in advance and stored in the common template storage unit 1 as storage means.

  In the registration process, when a fingerprint image indicating the user's fingerprint is acquired, a registration ID pattern X is generated from the user's fingerprint image by an ID pattern generation process (described later) using the common template information. In this authentication method, the registered ID pattern X is obtained by performing error correction encoding processing on the secret information S having a predetermined number of bits expressed in binary using a predetermined error correction encoding method. A check code C having the same code length can be generated.

Here, as an example of the error correction coding scheme, the error correction coding application means described in the document “A. Juels and M. Sudan ,:“ A fuzzy commitment scheme. ”ACM CCS, 1999” is used. Yes.
Detailed explanation of “A fuzzy commitment scheme” is omitted here, but in “A fuzzy commitment scheme”, auxiliary information is generated from registered data, and the verification information is slightly different from registered data using the auxiliary information. In this method, registered data is decoded from data using an error correction code. That is, the present invention is characterized in that the registration ID pattern X and the verification ID pattern X ′ are generated in the ID pattern generation processing using the common template information, and in order to correct a slight error during the verification processing, A biometric encryption method using an error correction coding applying means in the “A fuzzy commitment scheme” is proposed for these registered ID pattern X and verification ID pattern X ′.

  In the registration process, according to the means for applying the error correction encoding scheme in the “A fuzzy commitment scheme”, the secret information S is error correction encoded by a code word having an error correction capability of, for example, k bits (k is an arbitrary integer). As a result, a check code C having the same code length as that of the registered ID pattern X is generated and stored in a predetermined storage unit (not shown). Next, in this registration process, the exclusive logical sum of the registration ID pattern X and the check code C (hereinafter, the exclusive logical sum is indicated by a symbol in which “+” is entered in “◯”). (Step SP1), and the calculation result is stored in the storage server 2 as mask data M. In this way, in the registration process, the contents of the fingerprint information are converted into common template information, and the mask data M obtained by performing various processes to be described later is stored in the storage server 2. Is prevented from being stored in the storage server 2 as it is.

  In the collation process, when a fingerprint image indicating a fingerprint of a user who is an authentication target is acquired, the common template information used to generate the registration ID pattern X during the registration process is used, and the collation ID pattern is obtained from the fingerprint image by the ID pattern generation process. X ′ is generated. At this time, if the collation ID pattern X ′ and the registration ID pattern X are generated from the same user's fingerprint, a slight difference caused by finger misalignment during the registration process and the collation process can be limited. .

  In the verification process, after the mask data M is read from the storage server 2, the exclusive OR of the mask data M and the verification ID pattern X ′ is calculated to calculate the verification code C ′ (step SP2). The verification code C ′ is an exclusive OR of the verification ID pattern X ′, the registered ID pattern X, and the check code C, and the error part ε in the verification ID pattern X ′ and the registered ID pattern X It is also an exclusive OR with the check code C. The error portion ε indicates a difference in bit data between the verification ID pattern X ′ and the registered ID pattern X, and also indicates a difference in bit data between the verification code C ′ and the check code C. At this time, if the error portion ε is small enough to be restored, the secret information S can be restored.

  In practice, in the collation process, when the collation code C ′ is calculated, the collation code C ′ and the check code C are compared, and an error part ε indicating the difference between the collation code C ′ and the check code C is When the error correction capability is within the threshold k bits, the verification code C ′ is estimated to be a check code C, and the check code C is subjected to an error correction decoding process that is the reverse of the error correction encoding process. The secret information S is restored.

  Here, in the authentication method according to the present invention, the registered ID pattern X is not the fingerprint information itself, but the user's fingerprint information is changed by the ID pattern generation processing using the common template information, and is exclusive from the registered ID pattern X. Since the fingerprint data itself is not included in the ORed mask data M, it is difficult to guess the fingerprint information from the mask data M even if the inspection code C is leaked. It is possible to prevent the fingerprint information from being specified.

(2) Configuration of Authentication System Next, the authentication method described above will be described in detail using the authentication system 5 shown in FIG. 2 having a configuration different from that in FIG. In the case of this embodiment, the authentication system 5 includes a client terminal 6 and a server 7, and can transmit and receive data between the client terminal 6 and the server 7. Here, the common template information which is a feature of the present invention is stored in advance in a common template storage unit 8 provided in the server 7, and the ID pattern generation unit 10 of the client terminal 6 converts the common template information into the common template storage unit 8. The registered ID pattern X and the verification ID pattern X ′ can be generated by the ID pattern generation process using the common template information.

(2-1) Registration Process Here, the registration process will be described first. In this case, when the user's finger is placed on the fingerprint collecting surface of the fingerprint obtaining unit 11, the client terminal 6 can obtain a fingerprint image representing a ridge of the fingerprint on the fingerprint collecting surface. . The fingerprint image can be generated by, for example, an electrostatic method for generating a fingerprint image based on a charge amount difference between a crest (ridge) and a trough of a fingerprint generated when a finger is pressed against a semiconductor surface of the fingerprint collecting surface, or a fingerprint collecting surface. A heat-sensitive method that generates a fingerprint image by detecting the temperature difference between a mountain (body temperature) and a valley (air temperature) in contact with an electric field, and an electric field between a mountain (ridge) and a valley generated by a weak current flowing on the surface of the finger Various other methods such as an electric field method for generating a fingerprint image from a distribution pattern resulting from a difference in intensity may be used. The fingerprint acquisition unit 11 may acquire a fingerprint image in which a ridge of the fingerprint is clarified by imaging a user's fingerprint with an image sensor and performing predetermined image processing such as binarization. In short, as long as a fingerprint image clearly showing the end points and branch points of the fingerprint is obtained, various other image processing may be used.

  When the finger is placed on the fingerprint collecting surface and a fingerprint image composed of ridges representing the fingerprint is generated, the fingerprint acquisition unit 11 sends this as fingerprint image data B to the ID pattern generation unit 10. The ID pattern generation unit 10 serving as a generation unit and a conversion unit starts an ID pattern generation process, divides a fingerprint image generated based on the received fingerprint image data B into a predetermined number of areas, and forms a fingerprint. The fingerprint identification information for each area that analyzes the shape of the ridge can be generated for each area. Here, the ID pattern generation unit 10 can divide the fingerprint image into 85 areas, for example, with area numbers 0 to 84 in order.

  As shown in FIG. 3, each area ER that divides the fingerprint image has a circular shape and is arranged without a gap so that a part of the area overlaps each other, and all the fingerprint images are always located in any one of the areas ER. Are arranged to be. In addition, in the area segmentation area 15 composed of these areas ER, the minimum area ER is concentrically arranged in the central portion 16, and a plurality of areas ER are provided so as to surround the area ER of the central portion 16. Has been placed.

  Here, the area ER surrounding the area ER of the central portion 16 is arranged along a plurality of circular imaginary lines 17 selected so that the radius gradually increases as the distance from the central portion increases with the central portion 16 of the area division region 15 as the center. In addition, the center point 18 of each area ER is arranged on the circular virtual line 17 so as to overlap.

  In addition, the radius of the area ER is selected for each circular imaginary line 17 to be arranged, and the radius of the area ER arranged on the circular imaginary line 17 becomes closer to the circular imaginary line 17 that is separated from the central portion 16. Is selected to gradually increase. That is, the area ER adjacent to the area ER of the center portion 16 has a small radius, and the area ER arranged on the outermost circular virtual line 17 farthest from the center portion 16 is selected to have the largest radius. Yes.

  By the way, if the shape of the area ER that divides the fingerprint image here is a square shape, a corner is formed on the diagonal side from the center point of the area in each area. The distance to the boundary line is not constant, and the displacement tolerance in the oblique direction (angular direction) is increased. On the other hand, in the ID pattern generation unit 10, the distance from the center point 18 of the area ER to the boundary line with the adjacent area ER is constant by making the shape of the area ER for dividing the fingerprint image circular. The tolerance of fluctuation with respect to the positional deviation is made uniform.

  In addition, fingerprints generally have denser end points and branch points toward the center of the finger, and tend to have fewer end points and branch points toward the outside of the finger. Therefore, the ID pattern generation unit 10 reduces the radius of the area ER on the side of the central part 16 where the central part of the finger is easily arranged, and the area ER is denser as the distance from the central part 16 increases. The number of end points and branch points included in the area ER near 16 and the area ER far from the center 16 can be made uniform.

  Here, the ID pattern generation unit 10 analyzes the fingerprint image in which the ridges of the fingerprint are displayed, and determines the position where the ridge 21 is interrupted in the fingerprint image 20 as the end point 22 as shown in FIG. On the other hand, a position where one ridge 21 is branched and divided into a plurality of ridges 21 in the fingerprint image 20 can be determined as a branch point 23. The ID pattern generation unit 10 represents the determined end point 22 and branch point 23 in, for example, a binary system, and associates “0” with the determined end point 22 and “1” with the branch point 23. Yes.

  When the ID pattern generation unit 10 determines in which area ER the specified predetermined end point 22 or branch point 23 is located in the fingerprint image 20 obtained from the user's fingerprint, the ID pattern generation unit 10 Rather than using only the distance from the center point 18 to the end point 22 or the branch point 23, the relative distance with respect to the radius of the area ER is used as a matching threshold. Specifically, whether or not the end point 22 or the branch point 23 to be determined is included in the specific area ER is determined from the center point 18 of the area ER to the end point 22 or the branch point as shown in FIG. When d / r represented by the distance d to 23 and the radius r of the area ER is 1 or less, it is determined that the end point 22 or the branch point 23 is included in the area ER. In this way, the ID pattern generation unit 10 calculates d / r using the radius r of each area ER and the distance d from the center point 18 of the area ER to the end point 22 or the branch point 23 to be determined. Based on whether or not the result is 1 or less, it is determined in which area ER the end point 22 or the branch point 23 to be determined is included, and the area numbers 0 to 84 are determined. go.

  Next, as shown in FIG. 4, the ID pattern generation unit 10 can calculate angles θ <b> 1 and θ <b> 2 indicating in which direction each ridge 21 extends around the end point 22 and the branch point 23. Has been made. In this case, the ID pattern generation unit 10 uses, for example, the direction of the ridge 21 (characteristic portion) extending linearly from the end point 22 with the vertical line passing through the end point 22 and the branch point 23 in the fingerprint image 20 as a reference line y. Hereinafter, the angle (hereinafter referred to as the end point angle) θ1 formed by x1 and the reference line y and the direction of the ridge 21 (characteristic portion) linearly extending from the branch point 23 (hereinafter referred to as the branch point). An angle (hereinafter referred to as a branching point angle) θ2 formed by x2 and a reference line y can be calculated.

  In this way, the ID pattern generation unit 10 includes area information indicating the area ER having the end points 22 and the branch points 23 by area numbers 0 to 84, and attribute information indicating the end points 22 and the branch points 23 by attribute numbers 0 and 1. The area-specific fingerprint identification information including the angle information indicating the end point angle θ1 and the branch point angle θ2 can be generated for each area ER that divides the fingerprint image 20. In the above-described embodiment, the case where the vertical line is set as the reference line y passing through the end point 22 and the branch point 23 in the fingerprint image 20 as the reference line is described, but the present invention is not limited thereto. Various other direction lines such as a horizontal line passing through the end point 22 and the branch point 23 in the fingerprint image 20 may be used as the reference line y.

  In addition to such a configuration, as shown in FIG. 2, the ID pattern generation unit 10 reads out the common template information from the common template storage unit 8 provided in the server 7, and generates these based on the common template information. A plurality of fingerprint identification information by area consisting of area information, attribute information, and angle information is replaced with a corresponding common template. Here, as shown in FIG. 6, the common template information includes the fingerprint image 20 for the index number (in this case, the number from 0 to 5439 since 5440 common templates 30 are created in advance). Template area information 31 that represents the position of each area ER that divides the area with the area numbers 0 to 84, and the template attribute that represents the end points 22 and the branch points 23 of the ridges 21 in the fingerprint image 20 with the attribute numbers 0 and 1 Information 32 and template angle level information 33 representing all possible angles of the end point angle θ1 and the branch point angle θ2 of the ridge line 21 in the fingerprint image 20 in a predetermined angle range as a rough angle level. It is associated.

  Here, the template area information 31 is provided with area numbers of 0 to 84 corresponding to the number of areas obtained by dividing the fingerprint image 20 by the ID pattern generation unit 10. The template attribute information 32 represents the end point 22 and the branch point 23 in a binary system. For example, the end point 22 is represented as “0” and the branch point 23 is represented as “1”. Further, as shown in FIG. 7, the template angle level information 33 divides all angles of 360 degrees assumed by the end point angle θ1 and the branch point angle θ2 into 32 angle areas by equal angles, and 360 degrees is set to the angle level 0. It is classified into ˜31 levels (that is, the angle range of 1 level is 11.25 degrees).

  In the common template information composed of the plurality of common templates 30, two types of template attribute information 32 of the end point 22 and the branch point 23 are associated with the template area information 31 of area numbers 0 to 84, respectively. Template angle level information 33 of angle levels 0 to 31 is associated with the template attribute information 32. In other words, in the case of this embodiment, the common template information includes a total of 5440 items (index numbers 0 to 5439): area number (85) × attribute number (two types of end points and branch points) × angle level number (32 levels). ) Common template 30 is created in advance.

  Here, the ID pattern generation unit 10 selects one of the corresponding common templates 30 from the common template information based on the area information, attribute information, and angle information generated from the fingerprint image 20 indicating the user's fingerprint. The fingerprint identification information for each area is converted into the common template 30. When the fingerprint identification information for each area of the end point 22 consisting of, for example, the area number “0”, the attribute number “0”, and the end point angle “5 degrees” is obtained, the ID pattern generation unit 10 selects the area-specific fingerprint information from the common template information. Common template indicating the angle level “0 level” corresponding to the area number “0” and the attribute number “0” of the fingerprint identification information and including the end point angle “5 degrees” (that is, the common template having the index number “0”) ) 30 is selected, and the fingerprint identification information for each area is designated as the common template (template area information 31 is area number “0”, template attribute information 32 is attribute number “0”, and template angle level information 33 is angle level “0”. Level ”common template) 30.

  In addition, when the ID pattern generation unit 10 obtains fingerprint identification information for each area of the branch point 23 including, for example, the area number “84”, the attribute number “1”, and the branch point angle “350 degrees”, the common template A common template (ie, an angle level “31 level” that corresponds to the area number “84” and the attribute number “1” of the fingerprint identification information for each area and includes the branch point angle “350 degrees” from among the information) The common template (index number “5439”) 30 is selected, and the fingerprint identification information for each area is designated as this common template (template area information 31 is area number “84”, template attribute information 32 is attribute number “1”, template angle Level information 32 is converted into a common template 30) with an angle level of “31 levels”.

  In this way, the ID pattern generation unit 10 converts all of the obtained fingerprint identification information for each area into a common template 30, and a plurality of common templates (hereinafter referred to as “registration”) replaced with the fingerprint identification information for each area. A group (referred to as a time selection common template) can be temporarily stored as conversion information. Next, the ID pattern generation unit 10 identifies the selection common template at the time of registration in which the area number of the template area information is the same in the conversion information, and the level average of the template angle level information respectively associated with the same area number A value (hereinafter referred to as angle average) is calculated. When there are different template attribute information 32 for the end point 22 and the branch point 23 in the same area number, the angle average is calculated by combining the angle levels of the end point 22 and the branch point 23, and the attribute number of the template attribute information 32 is obtained. Regardless of (type of end point 22 and branch point 23), the angle average can be calculated for each area number.

  Next, the ID pattern generation unit 10 can classify the angle average into, for example, four areas 0 to 3 using an area determination table T100 as shown in FIG. In the case of this embodiment, the region determination table T100 has an angle average of 0 level or more and less than 8 levels (0 ≦ angle average <8) as region 0, and 8 levels or more and less than 16 levels (8 ≦ angle average <16). Region 1 is set to 16 or more and less than 24 levels (16 ≦ angle average <24), and region 2 is set to 24 or more and less than 32 levels (24 ≦ angle average <32).

  When the ID pattern generation unit 10 determines any one of the regions 0 to 3 based on the region determination table T100 from the angle average calculated for each area ER, each ID pattern generation unit 10 determines each of the regions based on the output bit conversion table T101 as illustrated in FIG. The output bit string associated with the areas 0 to 3 is read, and the areas 0 to 3 are converted into output bit strings.

  In practice, in this embodiment, the output bit conversion table T101 includes, for example, an output bit string “0000” in area 0, an output bit string “0001” in area 1, an output bit string “0010” in area 2, and an output bit string in area 3 “0011” is associated. In the output bit conversion table T101, the output bit string “1111” is associated with the case where there is no corresponding area. Thus, the ID pattern generation unit 10 associates one of the output bit strings corresponding to the output bit strings “0000”, “0001”, “0010”, and “0011” with the area ER in which the areas 0 to 3 are calculated. In addition, an output bit string “1111” is associated with an area ER in which there are no end points 22 and no branch points 23 and the areas 0 to 3 are not calculated, and the registration ID is a combination of these output bit strings in the order of area numbers 0 to 84. A pattern X is generated and sent to the mask data generation unit 40, and the ID pattern generation process is terminated.

  Next, a series of processing steps from receiving the above-described fingerprint image 20 to generating the registered ID pattern X in the ID pattern generation unit 10 will be summarized with reference to the schematic diagrams of FIGS. Here, for convenience of explanation, FIG. 10A shows that the shape of the fingerprint image 20 is divided by a plurality of areas ER, in order to clarify the boundary line between the areas ER. Is shown. 10A indicates the position of the end point 22 or the branch point 23. For example, the area ER1 has two end points 22 or 23, and the area ER2 has the end point 22 or the branch point 23. Indicates that there is none.

  These end points 22 or branch points 23 are replaced with a common template 30 composed of template area information 31, template attribute information 32, and template angle level information 33, and the angle average of the angle levels of the common template 30 for each area ER. Can be calculated. When the angle average is calculated for each of these areas ER, the ID pattern generation unit 10 converts the angle average into regions 0 to 3 based on the region determination table T100, and the angle average is calculated as shown in FIG. The areas 0 to 3 are associated with the calculated areas ER. In FIG. 10B, for example, the diagonal line of area ER2 indicates that there is no corresponding region 0 to 3 because there is no end point 22 or branch point 23 in area ER2, and no angle average is calculated. Show.

  Next, as shown in FIG. 10C, the ID pattern generation unit 10 converts the areas 0 to 3 into corresponding output bit strings based on the output bit conversion table T101, and determines an output bit string for each area ER. Go. Here, for example, the area ER that does not have the end point 22 or the branch point 23 and is not associated with the areas 0 to 3 is associated with the corresponding area 0 to 3 based on the output bit conversion table T101. The output bit string “1111” is associated. In this way, the ID pattern generation unit 10 associates the output bit strings with all the areas ER, and as shown in FIG. 10D, the registered ID pattern X in which these output bit strings are combined in the order of area numbers 0 to 84. Is transmitted to the mask data generation unit 40.

On the other hand, as shown in FIG. 2, the error correction encoding unit 41 receives the secret key S _k as the secret information according to application means of the error correction coding scheme in "A fuzzy commitment scheme", e.g., k bits ( The secret key _Sk is error-correction-encoded with a codeword having an error correction capability (k represents an arbitrary integer), and a check code C having the same code length as the registered ID pattern X is generated. The error correction coding unit 41 sends this check code C to the mask data generation unit 40 and stores it in a storage unit (not shown).

  The mask data generation unit 40 generates mask data M by calculating an exclusive OR of the registered ID pattern X and the check code C, and stores the mask data M in a storage medium 42 such as an IC (integrated circuit) card. Remember. Here, the registered ID pattern X is not generated directly from the user's fingerprint, but the fingerprint identification information obtained for each area ER of the fingerprint image 20 is replaced with the common template 30, and further obtained from the common template 30. Since it is composed of random bits converted into an output bit string based on the average angle for each area ER, the mask data M obtained by performing an exclusive OR with the registered ID pattern X also becomes random bits. As a result, it is difficult for the authentication system 5 to guess the user's fingerprint information (collection of fingerprint identification information for each area) from the mask data M, and it is possible to prevent the fingerprint information of each individual from being specified.

(2-2) Collation Processing Next, collation processing for authenticating whether or not the user to be authenticated is a registered regular user using the mask data M stored in the storage medium 42 will be described below. . In this case, as in the registration process, when the user's finger is placed on the fingerprint collection surface of the fingerprint acquisition unit 11, the client terminal 6 acquires the fingerprint image 20 representing the ridges of the fingerprint. It is sent to the ID pattern generator 10 as fingerprint image data B ′. As in the registration process, the ID pattern generation unit 10 includes, for each end point 22 and branch point 23 in the fingerprint image 20, area information indicating the area ER where the end point 22 or the branch point 23 is located by area number, and the end point The attribute information indicating 22 or the branch point 23 by an attribute number and the angle information indicating the end point angle θ1 or the branch point angle θ2 can be specified.

  That is, the ID pattern generation unit 10 prints fingerprints by using a plurality of circular areas (85 areas sequentially assigned with area numbers 0 to 84) ER as shown in FIG. The image 20 can be segmented, and the shape of the ridges 21 constituting the fingerprint can be analyzed for each area ER. In this case, the ID pattern generation unit 10 analyzes the fingerprint image 20 on which the fingerprint ridges 21 are displayed, and, as shown in FIG. On the other hand, the position where one ridge line 21 is branched into a plurality of ridge lines 21 in the fingerprint image 20 is determined as a branch point 23, and the determined end point 22 and branch point 23 are determined. Is expressed in binary, for example, so that “0” is associated with the determined end point 22 and “1” is associated with the branch point 23.

  Further, the ID pattern generation unit 10 has a collation threshold value d / r = 1 (d is a distance from the center point 18 of the area ER to the end point 22 or the branch point 23, and r is a radius of the area ER). Based on whether or not (FIG. 5), the area ER including the end point 22 and the branch point 23 is specified, and the end point 22 and the branch point 23 to be determined are determined in which area ER, The area numbers including the end points 22 and the branch points 23 are determined. As shown in FIG. 4, the ID pattern generation unit 10 can calculate the end point angle θ1 and the branch point angle θ2 with the vertical line passing through the end point 22 and the branch point 23 in the fingerprint image 20 as the reference line y. In this way, the ID pattern generation unit 10 performs the area information indicating the area ER where the end point 22 or the branch point 23 is located by the area numbers 0 to 84 and the end point 22 or the branch point 23 at the time of the matching process and the registration process. Can be generated for each area ER that divides the fingerprint image 20, each of which includes the attribute information indicating the attribute number 0 or 1 and the angle information indicating the end point angle θ1 and the branch point angle θ2. Has been made.

  In addition to this configuration, the ID pattern generation unit 10 reads the common template information from the common template storage unit 8 provided in the server 7 as in the registration process, and generates an area generated from the fingerprint image 20 indicating the user's fingerprint. Based on the information, attribute information, and angle information, one of the corresponding common templates 30 is selected from the common template information, and the fingerprint identification information for each area is converted into the common template 30, respectively.

  When the fingerprint identification information for each area of the end point 22 consisting of, for example, the area number “0”, the attribute number “0”, and the end point angle “5 degrees” is obtained, the ID pattern generation unit 10 selects the area-specific fingerprint information from the common template information. Common template indicating the angle level “0 level” corresponding to the area number “0” and the attribute number “0” of the fingerprint identification information and including the end point angle “5 degrees” (that is, the common template having the index number “0”) ) 30 is selected, and the fingerprint identification information for each area is designated as the common template (template area information 31 is area number “0”, template attribute information 32 is attribute number “0”, and template angle level information 33 is angle level “0”. Level ”common template) 30.

  In this way, the ID pattern generation unit 10 converts all of the fingerprint identification information for each area obtained during the matching process into the common template 30 and replaces the fingerprint identification information for each area with a plurality of common templates (hereinafter referred to as “fingerprint identification information for each area”). This group is called a collation selection common template), and can be temporarily stored as conversion information. Next, the ID pattern generation unit 10 specifies the selection common template at the time of matching in which the area number of the template area information 31 is the same in the conversion information, and the template angle level information 33 respectively associated with the same area number. A level average value (hereinafter referred to as an angle average) is calculated. Here, if there is different template attribute information 32 of the end point 22 and the branch point 23 in the same area number, the angle average of the end point 22 and the branch point 23 is collectively calculated, and the template attribute information 32 Regardless of the attribute number (the type of the end point 22 and the branch point 23), the angle average can be calculated for each area number.

  Next, the ID pattern generation unit 10 classifies the angle averages into, for example, four areas 0 to 3 based on the area determination table T100 as shown in FIG. 8, and then based on the output bit conversion table T101 as shown in FIG. The output bit strings associated with the areas 0 to 3 are read out, and the areas 0 to 3 are converted into output bit strings. Thus, based on the output bit conversion table T101, the ID pattern generation unit 10 associates the output bit strings with all the areas ER, and generates a collation ID pattern X ′ in which these output bit strings are combined in the order of area numbers from 0 to 84. This is sent to the collation unit 43 (FIG. 2). When the user at the registration process and the user at the verification process are the same person, the verification ID pattern X ′ and the registration ID pattern X generated at the registration process generally indicate the fingerprint position on the fingerprint collection surface. Due to fluctuations in information due to deviation or the like, the information may be slightly different.

The collation unit 43 reads the mask data M stored in the storage medium 42 during the registration process, takes an exclusive OR of the mask data M and the collation ID pattern X ′, and calculates a collation code C ′. Is sent to the error correction decoding unit 44. The error correction decoding unit 44 reads the check code C from a predetermined storage unit (not shown), compares the check code C ′ with the check code C, and determines the difference between the check code C ′ and the check code C. It is possible to determine whether or not the indicated error part ε is within a threshold value k bits within the range of error correction capability. As a result, when the error correction decoding unit 44 determines that the error part ε is within the threshold k bits within the range of the error correction capability, the error correction decoding unit 44 estimates that the verification code C ′ is the check code C, and The secret key S _k is restored by performing an error correction decoding process opposite to the error correction encoding process on the code C. On the other hand, when the error correction decoding unit 44 determines that the error portion ε is equal to or greater than the threshold k bits and is outside the range of the error correction capability, the error correction decoding unit 44 estimates that the verification code C ′ and the check code C are different, It can be determined that the user to be authenticated is not a registered regular user.

Error correction decoding unit 44, to restore the secret key S _k, and outputs this to the encryption processing unit 45. The encryption processing unit 45 receives a random number as a challenge code from the random number generation unit 47 of the server 7, generates a response code by encrypting the secret key _Sk with the challenge code, and sends this response code to the server 7. When the server 7 receives the response code from the determination unit 48, the server 7 decrypts the response code with the public key P _k , and issues the challenge code sent by itself and sent to the client terminal 6, and information obtained by decrypting the response code. The user is authenticated by determining whether or not they match.

(3) Operation and Effect In the above configuration, the client terminal 6 uses the area information indicating the area ER where the end point 22 or the branch point 23 is located in the fingerprint image 20 by the area number and the end point 22 or the branch point 23 as the attribute number. The area-specific fingerprint identification information including the attribute information indicating and the angle information indicating the end point angle θ1 or the branching point angle θ2 is generated for each area ER that divides the fingerprint image 20. Further, in the client terminal 6, the template area information 31 indicating the area ER for dividing the fingerprint image 20 by the area number, the template attribute information 32 indicating the end point 22 and the branch point 23 by the attribute number for each area ER, and the end point angle For each area, a plurality of common templates 30 including template angle level information 33 in which all assumed angles of θ1 and branching point angle θ2 are divided into predetermined angle ranges and indicated by rough angle levels. The common template 30 corresponding to the fingerprint identification information is selected, and the fingerprint identification information for each area is converted into the corresponding common template 30.

  As a result, the client terminal 6 does not use the fingerprint identification information for each area indicating the characteristics of the user's fingerprint, but uses the common template 30 instead of the fingerprint identification information for each area. X or the collation ID pattern X ′ can be generated, and thus it is difficult to guess the fingerprint identification information for each area from the registered ID pattern X or the collation ID pattern X ′.

  Further, at the time of registration processing, the client terminal 6 specifies a common template selected at the time of registration in which the area numbers of the template area information are the same, calculates an angle average for each area number, and calculates four areas 0 based on the area determination table T100. From 0 to 3, regions 0 to 3 corresponding to the angle average are determined. In the client terminal 6, each area ER associated with the areas 0 to 3 and each area ER without the end points 22 and the branch points 23 and not associated with the areas 0 to 3 are based on the output bit conversion table T101. A predetermined output bit string is associated with each other, and a registration ID pattern X is generated by combining these output bit strings in the order of area numbers from 0 to 84.

  As described above, since the client terminal 6 calculates the average angle for each area ER from the common template selected at the time of registration and converts it into an output bit string, the fingerprint identification information by area indicating the characteristics of the user's fingerprint is obtained. It can be further modified to make it more difficult to estimate the fingerprint identification information for each area from the registered ID pattern X, thereby preventing the fingerprint information of each individual from being specified.

  In addition, the client terminal 6 calculates the angle average for each area ER from the authentication-selected common template in which the area number of the template area information is the same as in the registration process even during the matching process, and from the area determination table T100 After determining the areas 0 to 3 corresponding to the angle average, each area ER associated with the areas 0 to 3 and each area ER not associated with the areas 0 to 3 are based on the output bit conversion table T101. Then, a predetermined output bit string is associated with each other, and a collation ID pattern X ′ is generated by combining these output bit strings in the order of area numbers from 0 to 84. Thus, the client terminal 6 can further modify the area-specific fingerprint identification information indicating the characteristics of the user's fingerprint even during the collation process, and further estimate the area-specific fingerprint identification information from the collation ID pattern X ′. This makes it difficult to prevent the fingerprint information of each individual from being specified.

  According to the above configuration, the fingerprint image 20 is divided into a plurality of areas ER, and the angle formed between the reference line y and the ridge line 21 having the end point 22 and the branch point 23 which are the characteristic portions of the fingerprint in the fingerprint image 20 Are generated for each area ER, and a plurality of area-specific fingerprint identification information is generated. In addition, the end point angle θ1 of the fingerprint identification information for each area is selected from a plurality of common templates in which all assumed angles formed by the fingerprint characteristic portion and the reference line y are represented for each area ER in a predetermined angle range unit. And a common template 30 having an angle level including the branch point angle θ2 is selected for each fingerprint identification information for each area, and the fingerprint identification information for each area is converted into the corresponding common template 30 and the matching ID pattern X and the verification ID pattern X ′ is generated. Thereby, in the authentication system 5, it is possible to perform collation processing while modifying the fingerprint identification information for each area showing the characteristics of the user's fingerprint, and the fingerprint identification information for each area from the registered ID pattern X or the collation ID pattern X ′. It is also difficult to guess, and it is possible to prevent the fingerprint information of each individual from being specified.

(4) Other Embodiments The present invention is not limited to the present embodiment, and various modifications can be made within the scope of the gist of the present invention. For example, in the above-described embodiment, the case where the authentication system 5 in which the client terminal 6 and the server 7 are separated from each other is applied as the authentication device. However, the present invention is not limited thereto, and the client terminal 6 An authentication apparatus in which the configuration and the configuration of the server 7 are provided in one housing may be applied, or various other configurations may be applied.

  Further, in the above-described embodiment, the angle average is calculated for each area ER of the registration common template and the matching selection common template, and the registration ID pattern X is calculated based on the region determination table T100 and the output bit conversion table T101. However, the present invention is not limited to this, and the registration-time selection common template and the matching-time selection common template are represented as bit strings as they are, which are registered ID pattern X and matching ID pattern. Alternatively, X ′ may be used as the registration ID pattern X and the collation ID pattern X ′, which are obtained by calculating the average angle for each area ER of the registration common template and the collation selection common template.

  Furthermore, in the above-described embodiment, the case where the application unit of the error correction encoding scheme in the “A fuzzy commitment scheme” is used as the error correction encoding process has been described, but the present invention is not limited to this, and other The means for applying the error correction coding method may be adopted. Further, Hough conversion processing may be performed on various data such as the check code C.

  Furthermore, in the above-described embodiment, the case where the angle average is classified into four regions has been described. However, the present invention is not limited to this, and various other regions such as three, five, six, etc. In this case, the output bit string may be associated with the output bit conversion table T101 according to the number of areas.

(5) Evaluation Experiment (5-1) Conditions for Secret Information Restoration First, conditions for secret information restoration when an evaluation experiment using the authentication method of the present invention is performed will be described. As shown in FIG. 11, the number of elements of the secret information S (secret key S _k ) is k, the number of elements of the check code C is g, and registered in the output bit string selected from each area information during the collation process. Let mt be the same number of output bit strings as in the process, and mf be the number of different output bit strings. Also, let e be the number of erasure errors when there is no minutiae (end point and branch point) that matches the area information. From here, the element error in the polynomial of the check code C is mf + e. Therefore, the relationship of the coincidence / mismatch relationship (1) shown in FIG. 11 is established. The conditional expression indicating the error correction capability is the error correction conditional expression (2) shown in FIG. When the number e of erasure errors is deleted from the coincidence / mismatch relational expression (1) and the error correction conditional expression (2), the evaluation expression (3) is obtained. It used for the evaluation experiment of the restoration rate.

(5-2) Evaluation Result As a secret information restoration experiment, the probability of correctly restoring the secret information S of the user himself / herself and each of other users (hereinafter referred to as a restoration rate) was confirmed. Further, as a comparison of the number of divisions, the restoration rate was confirmed when the angle information θ was divided into four (that is, four areas) and eight (that is, eight areas). FIG. 12 shows the used fingerprints and other simulation specifications. The collation threshold d / r (d is the distance from the center point 18 of the area ER to the end point 22 or the branch point 23, and r is the radius of the area ER) differs depending on the number of divisions of the angle information θ. In the experiment, a value was selected so that the restoration rate of others was comparable. In FIG. 12, NFIS2 of the feature quantity extraction algorithm is known software that extracts end points and branch points that are feature points from a fingerprint image.

  FIG. 13 shows the relationship between the number of secret information elements and the restoration rate when the angle information θ is divided into four (four regions) and eight (eight regions). In FIG. 13, as an example, when the Reed Solomon code on the GF (2 ^ 8) Galois extension field is used as the error correction code, when k = 7, that is, when the secret information amount is 56 bits, the person restoration rate is 90% when dividing into four As described above, a restoration rate of 5% or less has been achieved. In general, the information amount of secret information composed of k symbols on the GF (2m) Galois extension field is m × k bits.

  FIG. 14 shows the result of verifying the relationship between the false match rate (FMR) and the non-false match rate (FNMR) (ROC curve). From FIG. 14, it can be seen that, when the image is divided into eight, by subdividing the bit scale, it becomes weak against changes in the angle information θ due to the rotation of the fingerprint, and the identity restoration rate is reduced compared with the case of dividing into four. From the above, it has been found that increasing the number of divisions does not necessarily improve the accuracy, but the four divisions provide the best accuracy, and the user himself / herself can be distinguished from other users.

1,8 Common template storage 5 Authentication system (authentication device)
6 Client terminal 7 Server
10 Pattern ID generator (generation means, conversion means)
11 Fingerprint acquisition unit
40 Mask data generator

Claims (5)

Generating means for dividing a fingerprint image into a plurality of areas, and generating a plurality of fingerprint identification information for each area indicating an angle formed by a feature portion of the fingerprint in the fingerprint image and a reference line for each area;
Storage means for storing a plurality of common templates representing all the angles assumed by the characteristic part of the fingerprint and the reference line for each area in a predetermined angle range unit;
From the plurality of common templates, a common template of the angle range including the angle of the fingerprint identification information for each area is selected for each fingerprint identification information for each area, and each fingerprint identification information for each area corresponds to each. An authentication device comprising: conversion means for generating pattern information converted into the common template. The converting means includes
After converting the fingerprint identification information by area into the common template, the average value of the angular range of the common template is calculated for each area, and the pattern information represented by the average value is generated. The authentication device according to claim 1. The converting means includes
The authentication apparatus according to claim 2, wherein an output bit string associated in advance according to the average value is read, and the pattern information represented by the output bit string is generated. The authentication apparatus according to any one of claims 1 to 3, wherein when the pattern information is stored as registration information in a storage unit, an error correction encoding process is performed on the pattern information. A generating step of dividing the fingerprint image into a plurality of areas, and generating a plurality of fingerprint identification information for each area indicating the angle formed by the feature portion of the fingerprint in the fingerprint image and the reference line for each area by the generating unit When,
The angle of the fingerprint identification information for each area is selected from a plurality of common templates in which all the angles assumed by the characteristic part of the fingerprint and the reference line are expressed for each area in a predetermined angle range unit. A conversion step of selecting a common template of the included angle range for each of the fingerprint identification information for each area by a conversion unit, and generating pattern information obtained by converting each of the fingerprint identification information for each area into the corresponding common template; An authentication method comprising:

Images