JP6199470B1 - Signature authentication system - Google Patents

Abstract

In a signature authentication system, a registered personal signature and an input personal signature are registered even if the order of writing, the addition of signatures, and forgetting to write occur between the signature information for registration and the signature information to be authenticated. Reduce the probability of judging as a mismatch. In a signature authentication system in which signature information input from a terminal is authenticated by a signature authentication server, the authentication server associates a plurality of signature information input from an input unit in time series. At each sampling point of two signatures, an instantaneous value of the distance between the signatures is calculated, and an instantaneous value of the distance between the signatures is calculated for each of the two signatures of the plurality of signature information. The maximum value of the series obtained by taking a moving average with the sample width is calculated for each instantaneous value of the distance between signatures, and when at least one of the maximum values is larger than a pre-registered threshold, there is a difference in the order of writing. And a determination unit for detecting whether or not. [Selection] Figure 3

Description

  The present invention relates to a signature authentication system, and more specifically, generates a template signature from pre-registered signature information, and outputs an authentication result by comparing the signature information input for authentication with the template signature. The present invention relates to a signature authentication system.

  Conventionally, as a method for identifying an individual who operates an information terminal with respect to access from an information terminal, (1) information such as a personal identification number that can be known only by the individual, information that can specify an individual such as a credit card number, etc. A method of registering in advance and collating with information input at the time of access; (2) digitally registering information having a significant difference that can identify an individual such as a signature and a voiceprint; There are known a method of collating with a signature and a voiceprint, and (3) a method of digitizing personal biometric features such as fingerprints and retinas and registering them in advance and collating them with input fingerprints and retinas.

  Among them, in the authentication technique using a signature, a method for digitizing handwritten signature information registered in advance as a feature amount of vector information from the center of gravity of the entire signature and comparing it with the feature amount of the input signature information ( For example, refer to Patent Document 1), a method of collating a feature amount obtained by extracting and normalizing a feature point of a specific character from preregistered signature information with a feature amount of input signature information (for example, Patent Document 2) is known.

On the other hand, as an evaluation index for such recognition technology, Non-Patent Document 1 defines the following evaluation index.
Registration failure rate (FTE: Failure to Enroll) ・ The probability that the signature entered a certain number of times will fail if the registration conditions are not met. ・ FRR (False Rejection Rate) ・ The registered personal signature is entered. Probability of judging the identity of the person to be inconsistent Probability of judging that the registered identity signature (FAR: False Acceptance Rate) and the registered signature of the other person (same name) match CMR: Clone Match Rate) ・ The probability that another person imitates maliciously and determines that the entered signature is a match by mistake.

  According to this evaluation index, there is a need for an authentication technique that has a low registration failure rate (FTE), a principal rejection rate (FRR), a stranger acceptance rate (FAR), and a clone matching rate (CMR). However, when configuring the signature authentication system, there is a trade-off relationship that if FTE is set to be low, FAR and CMR are also high, and if FRR is set to be low, FAR and CMR are high. . Accordingly, there is a need for a signature authentication technique that satisfies each of these conflicting evaluation indices.

Japanese Patent No. 4603675 Japanese Patent No. 5912570

ISO / IEC TR 19795-3: 2007, Information technology-Biometric performance testing and reporting-Part 3: Modality-specific testing

  FIG. 1 shows an overall processing flow in a conventional signature authentication system. As an example, a system will be described in which the signature authentication method described in Patent Document 1 is applied and the signature information input by the electronic pen on the display on the tablet terminal is the target. In this system, position information of the tip of the electronic pen on the display plane and pressure (writing pressure) information on the display plane of the electronic pen are handled as signature information. The signature authentication method calculates a feature quantity from pre-registered signature information, creates a template signature and registers it, calculates a feature quantity from the input signature information, It is roughly divided into an authentication processing flow for outputting an authentication result in comparison with a signature.

  In the registration process flow, the user's signature is input to the tablet terminal a plurality of times (S11). Even in the handwriting of the person himself, a certain amount of error always occurs, so registration signature information (hereinafter referred to as registration signature information) is acquired, for example, about 3 to 5 times. Even if they are the same person, the position of writing a signature on the display plane is different, and the size of characters is also different. Therefore, the signature area that defines the entire signature is normalized (S12).

  Next, in order to associate the signature information for registration with the signature information input for authentication (hereinafter referred to as authentication target signature information), the handwriting of the registration signature information is changed in time series. Sampling. For each sampling point, position information and writing pressure information in the normalized signature area are acquired (S13). This is because the registration signature information and the authentication target signature information are associated with each other as described below. Further, for comparison between the registration signature information and the authentication target signature information, vector information and writing pressure information from the center of gravity of the entire signature are digitized as feature quantities (S14).

  Such feature amount calculation is performed for registration signature information input a plurality of times, and registration signature information having a specific feature amount is excluded under a certain condition (S15). Is registered as a template signature (S16).

  In the authentication processing flow, when a signature input from a tablet terminal for authentication is acquired as authentication target signature information (S21), a signature area that defines the entire signature is normalized (S22). The handwriting of the authentication target signature information is sampled in time series, and the position information and the writing pressure information in the normalized signature area are acquired for each sampling point (S23). Furthermore, vector information and pen pressure information from the center of gravity of the entire signature are digitized as feature quantities (S24).

  Next, DP (Dynamic Programming) matching is performed between the registered template signature handwriting and the authentication target signature information handwriting to associate the handwriting (S25). For example, the vector information difference between one sampling point of the template signature and the sampling point of the corresponding authentication target signature information is accumulated over the entire signature, and DP matching is performed so that the total value is minimized. Whether the template signature and the authentication target signature information match or not is determined based on whether or not the total value of the differences is equal to or less than a predetermined threshold (S26). In addition, other feature amounts are extracted and compared from the sampling information of the template signature and the authentication target signature information, and compared with the threshold determination described above, the matching / non-matching determination accuracy is improved.

  In the signature authentication method, as described above, it is desirable to set the principal rejection rate (FRR) to be low. However, in the authentication processing flow, only the comparison of the feature amount composed of vector information and writing pressure information results in a high other-person acceptance rate (FAR) and clone matching rate (CMR) as described below. was there.

  Conventionally, in the registration processing flow, if the signature of the person is inputted a plurality of times (for example, three times) and the feature amounts of the three registration signature information are within a predetermined range, these three registration signature information Was registered as a template signature. However, even if it is a signature entered by the same person, if you change the order of writing, if you forget to write part of the font and add it later, you may forget to write part of the font. It happened not a little.

  As described above, in order to correlate the registration signature information with the authentication target signature information, in order to sample the handwriting in time series, when there is a mistake in the order of writing, writing, or forgetting to write, Even if the calculated feature amount is within a predetermined range among a plurality of registered template signatures, the difference becomes large.

  When the authentication process flow is executed with such a template signature set so that the probability (FRR) of judging that the inputted personal signature is not matched is low, the signature inputted maliciously by another person is mistakenly matched. The determination probability (CMR) becomes high. Conversely, if the authentication process flow is executed with the CMR set to be low, the FRR will be high.

  It is an object of the present invention to register a registered personal signature and an inputted personal signature even when a writing order is incorrect, an additional writing, or a forgetting writing occurs between the registration signature information and the authentication target signature information. It is an object of the present invention to provide a signature authentication system that can reduce the probability of determining a mismatch.

In order to achieve such an object, the present invention provides a method for detecting a writing order error or the like executed by a determination unit of a signature authentication server in a signature authentication system, wherein the determination unit includes the signature authentication. For a plurality of signature information input from the input unit of the server, for each sampling point (s = 1,..., S) of two signatures (S ₁ , S ₂ ) associated in time series, The instantaneous value of the distance between signatures from the xy coordinates of the two signatures with the index that associates the two signatures as (i (s), j (s))

A step of calculating an instantaneous value of a distance between signatures for each of two signatures of the plurality of signature information, and a series obtained by taking a moving average with a sample width registered in advance

A step of calculating a maximum value for each of the instantaneous values of distances between signatures, and when at least one of the maximum values is greater than a pre-registered threshold (C _seq )

And a step of detecting that the writing order is incorrect.

  As described above, according to the present invention, the instantaneous value of the inter-signature distance is calculated for each sampling point of two signatures associated in time series, and compared with a threshold value. Thus, it is possible to detect writing and forgetting to write, and to reduce the probability that the registered personal signature and the input personal signature are determined to be inconsistent.

It is a figure which shows the whole processing flow in the conventional signature authentication system. It is a figure which shows the signature authentication system concerning one Embodiment of this invention. It is a block diagram which shows the signature authentication server concerning one Embodiment of this invention. It is a schematic diagram for demonstrating the distance between signatures. It is a figure which shows the result of having unified the signature length when there is no writing order difference. It is a figure which shows the result of having performed the unification process of signature length when there exists a writing order difference.

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

  FIG. 2 shows a signature authentication system according to an embodiment of the present invention. For example, the case where signature authentication is used in place of conventional seal stamp authentication in banking operations will be described. A branch terminal 101 installed in the store 100 of each branch of the bank is provided with a tablet terminal 102 and an electronic pen 103. When a so-called “net banking” service is used in the general home 110, the user's tablet terminal 112 and the electronic pen 113 are used. The tablet terminals 102 and 112 are connected to the bank system 130 via the branch office terminal 101 or the router 111 and the network 120.

  The bank system 130 includes a front-end server 131 for communicating with the branch terminal 101 and providing the tablet terminal 112 with a web application for the net banking service. The signature authentication server 132 connected to the front-end server 131 includes authentication target signature information input from the tablet terminal 102 via the branch terminal 101 and authentication target signature information input from the tablet terminal 112 via the web application. Authenticate. If the signature authentication is successful, communication is performed between the front-end server 131 and the application server 133 (for example, a billing system), and a service such as a transfer is provided.

  FIG. 3 shows a signature authentication server according to an embodiment of the present invention. The signature authentication server 200 includes an input unit 201 that captures registration signature information and authentication target signature information from the front-end server, a preprocessing unit 202 that normalizes and samples the signature area of each signature information, and the entire signature. A feature amount calculation unit 203 that calculates vector information and writing pressure information from the center of gravity position as feature amounts. The feature amount calculation unit 203 registers the template signature calculated from the registration signature information in the signature database 211.

  The determination unit 204 associates the handwriting between the handwriting of the template signature registered in the signature database 211 and the handwriting of the authentication target signature information obtained from the feature amount calculation unit 203, and Match or mismatch is determined. The determination result is returned to the front end server via the output unit 205. The determination unit 204 determines match or mismatch according to the determination conditions registered in the authentication rule database 212.

(Unification of signature length)
In such a configuration, the determination unit 204 of the signature authentication server 200 performs DP matching on the registration signature information and authentication target signature information captured by the input unit 201 by unifying the signature length, Perform handwriting association. For example, in the registration processing flow, the two signatures S ₁ and S ₂ inputted as registration signature information are different in the length of the entire signature and the number of sampling points even if they are the same person. Therefore, using a technique called Dynamic Time Warping, the two signatures are associated in time series so that they overlap as much as possible. Specifically, the signature handwriting input from the tablet terminal is sampled in time series, and for each sampling point, the signature is as follows as the x and y coordinates from the reference point on the display plane. Stored in the database 211.
Signature data of signature S ₁ : (x _i (t), y _i (t)), t = 1,. . . , T _i
Signature data of signature S ₂ : (x _j (t), y _j (t)), t = 1,. . . , T _j
An index for associating _two signatures S ₁ and S ₂ is (i (s), j (s)), s = 1,. . . , S. The distance between the point i (s) of the signature S _{1 and} the point j (s) of the signature S ₂ is d (i (s), j (s)). Here, the association is formulated as follows, with the constraint of keeping the time order and associating all points.

i (1) = j (1) = 1
i (s) ≦ i (s + 1) ≦ i (s) +1
j (s) ≦ j (s + 1) ≦ j (s) +1
i (S) = Ti, j (S) = Tj
Corresponding sampling points of two signatures in time series, distance between two signatures at each sampling point

(I (s), j (s)) is determined so that the cumulative value of is minimized. That is,

It is sufficient to determine s to be such that DP matching is applied.

(Judgment of difference in shape)
In the registration processing flow described above, for example, the user inputs three signatures S ₁ , S ₂ , S _3, and the inter-signature distances D _xy (S ₁ , S ₂ ), D _xy (S ₂ , S _3). ), D _xy (S ₃ , S ₁ )
Calculate If any one of the threshold values C _xy exceeds the predetermined threshold value C _xy , the template signature is not registered, and the user is again required to obtain the registration signature information. Prompt for input.

The inter-signature distance will be described with reference to FIG. FIG. 4A shows the three inter-signature distances calculated as described above on a plane. The three vertices of the triangle represent the three signatures S ₁ , S ₂ , S ₃ , and each side corresponds to the inter-signature distance. The smaller this triangle, the closer the three signatures are. In the authentication processing flow, for the signature S ₄ input as the authentication target signature information,
D _xy (S ₁ , S ₄ ), D _xy (S ₂ , S ₄ ), D _xy (S ₃ , S ₄ )
If any one is within the range of the threshold value β, it is determined that the handwriting of the template signature matches the handwriting of the authentication target signature information. Here, the threshold value β is an inter-signature distance including the other two signatures when a circle is drawn around one signature.

However, even if it is a signature entered by the same person, if you change the order of writing, if you forget to write part of the font and add it later, if you forget to write part of the font, The distance between the three signatures of the signature information for registration becomes longer, as shown in FIG. For example, when the writing order is different in the signature S ₁ , the inter-signature distances D _xy (S ₁ , S ₂ ) and D _xy (S ₃ , S ₁ ) become longer. As described above, when the threshold value β is set, the threshold value β becomes larger than that in FIG. That is, a plurality of signatures registered by the same person are in the same writing order, and a threshold β that is larger than the distance between signatures when the signature is signed without forgetting to write, adding, etc. is set, and in the authentication processing flow, Others acceptance rate (FAR), clone matching rate (CMR) will become high.

(Detection of wrong order)
Therefore, it is detected that the writing order is wrong, the writing is overwritten, or the writing is forgotten. In the registration processing flow, registration is not accepted as in the case of the shape difference. FIG. 5 shows the result of performing the signature length unification process when there is no difference in the writing order. The broken line in FIG. 5A is one of the template signatures (S _t ) that is the target of handwriting association, and the solid line is the authentication target signature information (S _in1 ). FIG. 5B shows the result of calculating D _xy (S _t , S _in1 ) described above by performing DP matching on two signatures. The horizontal axis represents time, and the vertical axis represents the distance between two signatures at each sampling point.

FIG. 6 shows the result of performing the signature length unification process when there is a difference in the order of writing. The broken line in FIG. 6A is one of the template signatures (S _t ) that is the target of handwriting association, and the solid line is the authentication target signature information (S _in2 ). FIG. 6B shows the result of calculating D _xy (S _t , S _in2 ) described above by performing DP matching on two signatures. In the registration signature information (S _in2 ), as shown in Fig. 6 (a), forgetting to write the first character "Near" Shin Niu (X) and finishing writing the third character "Bold" After the point is written, the fourth character “Buro” is starting to be written.

As shown in FIG. 6B, when such a writing order difference occurs, the inter-signature distance at the sampling point takes a large value instantaneously. In determining the shape difference, since it is an instantaneous distance difference, there is no large difference in the difference in the accumulated value of the distance between signatures, and the determination of D _xy is not greatly affected. That is, in the conventional method, it was not possible to detect a writing order difference or the like.

Therefore, after unifying the signature lengths of the two signatures S _i and S _j input in the authentication processing flow, the distance between the sampling points of the two signatures associated in time series is set to x , Y coordinate. This distance is called the instantaneous value of the distance between signatures, and in the two signatures Si, Sj,

Is calculated. When the maximum value of a series obtained by taking a moving average with a preset sample width is larger than a preset threshold C _seq (dotted line in FIG. 6B)

It is determined that there was an incorrect writing order. At this time, for example, the tablet terminals 102 and 112 display that the writing order is wrong, and prompt the user to re-enter the signature.

In the registration process flow, the instantaneous value of the distance between the signatures of the three input signatures S ₁ , S ₂ , S ₃ is calculated. If the maximum value of a series obtained by taking a moving average with a preset sample width is larger than a preset threshold value C _seq , registration is not accepted and the user is prompted to input a signature again.

(Judgment such as wrong writing order)
In the above-described detection of a writing order difference or the like, DP matching is performed, and the instantaneous value of the inter-signature distance at each sampling point is compared with a preset threshold value. However, even if the signature has no writing order difference, or the signature of the person himself is very similar, as a result of DP matching, the instantaneous value for each sampling point is instantaneously May jump up. Therefore, when the inter-signature distance for each sampling point is compared with the threshold value, there is a possibility that the writing order is mistakenly detected even though the writing order is not wrong.

  Therefore, paying attention to the fact that the instantaneous value increases at a fixed number of consecutive sampling points when the writing order is wrong, the writing is overwritten, or forgetting to write, the value obtained by taking a moving average with a preset sample width is used. I decided to compare it with the threshold. Thereby, it is possible to suppress erroneous detection of a writing order or the like.

  Note that the sample width for taking the moving average is set to the time the electronic pen is moving (usually 100 to 200 milliseconds) when the writing order is incorrect, the writing is overwritten, or writing is forgotten. It is possible to improve the detection accuracy such as wrong writing order.

DESCRIPTION OF SYMBOLS 100 Store 101 Branch terminal 102,112 Tablet terminal 103,113 Electronic pen 110 General household 111 Router 120 Network 130 Bank system 131 Front end server 132,200 Signature authentication server 133 Application server 211 Signature database 212 Authentication rule database

Claims (8)

In a method for detecting a writing order error or the like executed by the determination unit of the signature authentication server in the signature authentication system
The determination unit performs sampling points (s = 1,... ) Of two signatures (S ₁ , S ₂ ) associated in time series with respect to a plurality of pieces of signature information input from the input unit of the signature authentication server . , S) , and the index for associating the two signatures is (i (s), j (s)), and the instantaneous value of the distance between the signatures from the xy coordinates of the two signatures

Calculating an instantaneous value of an inter-signature distance for each of two signatures of the plurality of signature information; and
A series in which a moving average is taken with a pre-registered sample width

Calculating a maximum value for each instantaneous value of the distance between signatures;
When at least one of the maximum values is greater than a pre-registered threshold (C _seq )

And a step of detecting that there is a writing order error.   The method according to claim 1, wherein the plurality of signature information uses a plurality of registration signature information input from the input unit in a registration processing flow for registering a template signature for authentication.   The determination unit outputs information indicating that registration signature information in which a writing order difference or the like has been detected is rejected when a writing order difference or the like is detected in the registration processing flow. Item 3. The method according to Item 2.   The method according to claim 2, wherein the plurality of pieces of signature information further uses authentication target signature information input from the input unit in an authentication processing flow. A computer-executable program for causing a computer to execute the steps according to any one of claims 1 to 4 . In a signature authentication system that authenticates signature information input from a terminal in a signature authentication server, the signature authentication server includes:
An input unit for acquiring the signature information input from the terminal;
For a plurality of signature information that is input from the input unit, time series association being two signatures are (S _1, S ₂₎ sampling point (s = 1, ..., S ) for each, the The instantaneous value of the distance between signatures from the xy coordinates of the two signatures, where the index that associates the two signatures is (i (s), j (s))

A series in which an instantaneous value of a distance between signatures is calculated for each of two signatures of the plurality of signature information, and a moving average is calculated with a sample width registered in advance

When the maximum value of each is calculated for the instantaneous value of the distance between the signatures, and at least one of the maximum values is greater than a pre-registered threshold (C _seq )

And a determination unit for detecting that the writing order is wrong. In a method for detecting a writing order error or the like executed by the determination unit of the signature authentication server in the signature authentication system,
The determination unit sets the sampling points (s = 1,..., S) of two signatures (S ₁ , S ₂ ) in time series for a plurality of pieces of signature information input from the input unit of the signature authentication server. And the distance between the two signatures at each sampling point from the xy coordinates of the two signatures.

Index (i (s), j (s)) that associates the two signatures so that the cumulative value of

Determining s such that
The determination unit determines an instantaneous value of a distance between signatures for each sampling point of two signatures associated in time series.

Calculating an instantaneous value of an inter-signature distance for each of two signatures of the plurality of signature information; and
A series in which a moving average is taken with a pre-registered sample width

Calculating a maximum value for each instantaneous value of the distance between signatures;
When at least one of the maximum values is greater than a pre-registered threshold (C _seq )

And a step of detecting that there is a writing order error. In a signature authentication system that authenticates signature information input from a terminal in a signature authentication server, the signature authentication server includes:
An input unit for acquiring the signature information input from the terminal;
The sampling points (s = 1,..., S) of two signatures (S ₁ , S ₂ ) are associated in time series with respect to a plurality of signature information input from the input unit, and the 2 The distance between two signatures at each sampling point from the xy coordinates of one signature

Index (i (s), j (s)) that associates the two signatures so that the cumulative value of

Determine s,
Instantaneous value of distance between signatures for each sampling point of two signatures associated in time series

And calculating an instantaneous value of the inter-signature distance for each of the two signatures of the plurality of signature information,
A series in which a moving average is taken with a pre-registered sample width

For each instantaneous value of the distance between signatures,
When at least one of the maximum values is greater than a pre-registered threshold (C _seq )

And a determination unit for detecting that the writing order is wrong.

Images