KR101056307B1 - Multiline Reliability Support Hybrid Online Signature Verification System and Its Method for Various Applications - Google Patents

Abstract

Disclosed are a multi-criteria reliability support hybrid online signature verification system and method for various applications. The multi-criteria reliability support hybrid online signature verification system includes a feature-based signature verification module that calculates a distance calculation value of feature-based signature verification, which is a distance between a vector extracted from a reference signature and a vector extracted from an input signature; A function-based signature verification module for calculating a distance calculation value of a function-based signature verification, which is a distance obtained by accumulating the difference between the reference signature and the input signature according to an input time flow of the reference signature and the input signature; And a signature verification result fusion module for determining the authenticity of the input signature by fusing the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification in series or in parallel.

Signature Verification, Multi-criteria Reliability, Feature-Based Signature Verification, Function-Based Signature Verification, Data Mining

Description

HYBRID ON-LINE SIGNATURE VERIFICATION SYSTEM SUPPORTING MULTI-CONFIDENTIAL LEVELS AND METHOD FOR THEREOF}

Embodiments of the present invention relate to an on-line signature verification system, and more particularly, to a hybrid on-line signature verification system and method for supporting multi-criterion signature verification reliability incorporating a feature-based signature verification technique and a function-based signature verification technique. will be.

As the related technology of digital pens or tablelets develops, the fields and applications requiring online signature verification (bank transactions, card transactions, electronic document settlement, etc.) are greatly increasing. In order to verify the identity of the user, for online signature verification purposes, the digital signature or a tablelet is used to determine the authenticity of the signature entered by the person who entered the signature. Therefore, the accuracy and promptness of signature verification is important in online signature verification.

There are two types of recognition errors that can occur in signature verification: false acceptance errors and false rejection errors. If the reliability criteria of signature verification are set high to reduce the false acceptance ratio (FAR) for the purpose of improving the accuracy of signature verification, the time required for signature verification generally increases. In addition, since the false rejection ratio (FRR) increases, the user's true signature is classified as a fake signature, causing inconvenience to the user. Using the same reliability criterion in multiple applications using online signature verification results in an unnecessarily high FRR for applications requiring relatively low reliability, and an unnecessarily high FAR for applications requiring relatively high reliability. It exists.

Therefore, in order to support various signature verification applications, online signature verification system with multi-criteria reliability support and systematic multi-criteria reliability setup method are required to support the reliability of signature verification according to the importance of each application service and the complexity of user signatures. Do.

The methods used in the existing online signature verification system can be largely divided into feature-based signature verification and function-based signature verification. Feature-based signature verification is a method of determining the authenticity of an input signature based on the distance in multidimensional space between a holistic vector representing a reference signature and a holistic vector representing an input signature. It is a method of determining the authenticity of an input signature based on the distance accumulated between the results of local feature comparison between the reference signature and the input signature over the passage of the signature time. Hybrid systems exist that combine the two techniques to improve the speed and accuracy of signature verification.

An embodiment of the present invention is a feature-based signature verification technique and a function-based signature that allows a system administrator and a user to dynamically select the reliability of signature verification according to the service content and importance of an application requiring online signature verification and the complexity of the user signature. The present invention provides a hybrid online signature verification system and method for supporting multi-standard signature verification reliability incorporating signature verification techniques.

One embodiment of the present invention provides a hybrid on-line signature verification system and method for establishing an ideal multi-criteria reliability using data mining technology by utilizing actual signature verification data using a signature verification input tool.

The multi-criteria reliability support hybrid online signature verification system according to an embodiment of the present invention is a feature-based signature verification module that calculates a distance calculation value of feature-based signature verification, which is a distance between a vector extracted from a reference signature and a vector extracted from an input signature. ; A function-based signature verification module for calculating a distance calculation value of a function-based signature verification, which is a distance obtained by accumulating the difference between the reference signature and the input signature according to an input time flow of the reference signature and the input signature; And a signature verification result fusion module for determining the authenticity of the input signature by fusing the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification in series or in parallel.

In one embodiment of the present invention, the signature verification result fusion module performs a first verification process that compares the distance calculation value of the feature-based signature verification with a feature-based acceptance criterion which is one criterion for determining the authenticity of the input signature. After that, a second verification process is performed to compare the distance calculation value of the function-based signature verification with a function-based acceptance criterion, which is another criterion for determining the authenticity of the input signature, according to the result of the first verification process.

In one embodiment of the present invention, the signature verification result fusion module uses the FAR corresponding to the distance calculation value of the feature-based signature verification and the FAR corresponding to the distance calculation value of the function-based signature verification. Similarity is calculated and the authenticity of the input signature is determined according to the calculated similarity.

The multi-criteria reliability support hybrid online signature verification method according to an embodiment of the present invention is a multi-criteria reliability support hybrid of a hybrid online signature verification system including a feature-based signature verification module, a function-based signature verification module, and a signature verification result fusion module. An online signature verification method, comprising: calculating a distance calculation value of feature-based signature verification, which is a distance between a vector extracted from a reference signature and a vector extracted from an input signature in the feature-based signature verification module; Calculating, by the function-based signature verification module, a distance calculation value of a function-based signature verification, which is a distance obtained by accumulating the difference between the reference signature and the input signature according to an input time flow of the reference signature and the input signature; And determining the authenticity of the input signature by fusing the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification in series or in parallel in the signature verification result fusion module.

In the embodiment of the present invention, the authenticity of the input signature may be compared with a feature-based acceptance criterion, which is a criterion for determining the authenticity of the signature of the input signature. If the distance calculation value of the feature-based signature verification is within the feature-based acceptance criterion, comparing the distance calculation value of the function-based signature verification with a function-based acceptance criterion which is another criterion for determining the authenticity of the input signature; And determining the input signature as a true signature if the distance calculation value of the based signature verification is within the function-based acceptance criterion.

In an embodiment of the present invention, the authenticity of the input signature may include calculating FRR and FAR for the reference signature using data mining, and authenticating the authenticity of the input signature according to the calculated FRR and FAR. Setting a threshold value, which is a criterion for determining, and determining the authenticity of the input signature by comparing the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification with the set threshold value. .

The calculating of the FRR and the FAR for the reference signature may include calculating a maximum true signature distance, which is the maximum distance among the distances between the reference signatures, and calculating a minimum true signature distance, which is the minimum distance among the distances between the reference signatures. Calculating a minimum garish name distance, which is the minimum distance between the reference name and the goji name forging the reference name, and calculating the minimum signature distance ratio, which is the ratio of the maximum signature distance and the minimum signature distance. Calculating a minimum garish distance ratio which is a ratio of the maximum true signature distance and the minimum garish name distance; calculating a FRR using the minimum true signature distance ratio; Includes steps to calculate FAR using proportions.

In an embodiment of the present invention, the authenticity of the input signature may include extracting a FAR corresponding to a distance calculated value of the feature-based signature verification and a FAR corresponding to a distance calculated value of the function-based signature verification. Calculating a similarity between the reference signature and the input signature using the FAR corresponding to the distance calculation value of the feature-based signature verification and the FAR corresponding to the distance calculation value of the function-based signature verification; Determining the authenticity of the input signature according to the similarity.

According to an embodiment of the present invention, by appropriately selecting the appropriate reliability criteria according to the importance of the application service and the complexity of the user signature to perform the user authentication, it is possible to secure the accuracy of the appropriate signature verification for each application of the online signature verification. Alternatively, the signature verification method and reliability criteria can be determined dynamically according to the signature input environment.

According to an embodiment of the present invention, an application for supporting user authentication by selecting a signature verification reliability for each user according to the complexity of the user signature, an application for selecting a user authentication reliability according to service type / content and service importance, It is possible to increase the reliability of signature verification automatically or in a user-selective manner in applications that provide a low level of signature verification reliability in a pre-registered environment and an automatic adjustment function for user authentication with a high signature verification reliability in an environment where the user does not pre-register. It can be dynamically adjusted to provide signature verification security suitable for the environment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited or limited by the embodiments. Like reference numerals in the drawings denote like elements.

1 illustrates an internal configuration of a hybrid online signature verification system according to an embodiment of the present invention. Referring to FIG. 1, a registration process of a reference name and a verification process of an input signature in a hybrid online signature verification system according to an embodiment of the present invention will be described.

The signature input module 100 performs a function of receiving a signature, and refers to an input device and a driving software module that receive a signature by a digital pen or a tablelet.

The preprocessing module 110 normalizes the input signature information according to the signature input time or the size of the signature through the preprocessing process.

The feature information extraction module 120 performs a function of extracting feature information, which may be mainly used to determine the authenticity of the signature, from the original signature information. Here, the feature information may be empirically selected by the system developer by referring to the feature information used in the field of character recognition and image recognition, and the distinguishing feature having a high discrimination ability may be selected by using the attribute selection technique of the data mining technique.

The feature-based signature registration module 130 stores the extracted values (ie, feature information) of the attribute information used for feature-based signature verification in the feature-based reference signature DB 150, and the function-based signature registration module 140 functions. Extraction values (ie, feature information) of the attribute information used for the base signature verification are stored in the function-based reference signature DB 160.

The feature-based signature verification module 170 constructs a multidimensional vector using the overall form of the signature and the overall speed and time as feature information to calculate a distance between the reference signature and the input signature.

The function-based signature verification module 180 calculates the cumulative distance of the difference according to the input time flow of the reference signature and the input signature using a dynamic time warping distance function calculation formula.

The signature verification result fusion module 190 performs a function of fusing the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification according to an appropriate form and determining the authenticity of the input signature. The signature verification result fusion module 190 may fuse the result values of the feature-based signature verification module 170 and the function-based signature verification module 180 in series or in parallel to support various applications, even in the case of serial fusion. It supports multi-standard signature verification reliability to secure the security of signature verification suitable for application.

2 is a diagram illustrating a signature verification process through serial fusion in a hybrid online signature verification system according to an embodiment of the present invention.

Hybrid online signature verification system according to an embodiment of the present invention calculates the distance (D _feature , D _function ) between the reference name and the input signature through the feature-based signature verification module 170 and the function-based signature verification module 180 (S201).

In detail, the feature-based signature verification module 170 retrieves k reference signatures (k = 5) by the signer ID from the feature-based reference name DB 150 to find the registered reference name of the user, Through 1, the distance (D _feature ) between the feature information of the reference signature and the feature information of the verification input signature is calculated.

Here, k denotes each reference signature (k = 1 to 5), R _{k , n} denote n-th feature information of the k-th reference signature, and C _{n denote} n-th feature information of the input signature. The CD is calculated using the Canberra distance as shown in Equation 2 as a method for calculating the distance between the feature information of the reference signature and the feature information of the input signature. In this case, the reason why the minimum value of the distance between the feature information of the various reference signatures and the feature information of the input signature is to compare the feature information of the reference signature with the feature information of the closest reference signature. The method of calculating Canberra distance has the effect of normalizing feature information having different ranges and units of values.

The function-based signature verification module 180 determines the similarity between the registered reference name of the user and the input signature input for verification, and the characteristics of the reference signatures in the function-based reference name DB 160 to determine the similarity. Compute the cumulative distance between the information. The distance calculation may use a dynamic time warping algorithm, which is a kind of dynamic program. A method of calculating a cumulative distance (D _function ) between the feature information of the reference signatures and the feature information of the input signature is shown in Equation 3 below.

Here, M is the number of local feature information (M = 4) used in the function-based signature verification, and DTW (R _{k , m} , C _m ) is the m-th feature of the k-th reference signature (k = 1-5). The distance between the information R _{k , m} and the m-th characteristic information C _m of the input signature is obtained through a dynamic time warping algorithm. In this case, the reason why the minimum value of the distance between the feature information of each reference signature and the feature information of the input signature is taken is to compare the feature information of the input signature with the feature information of the closest reference signature. And, the distance obtained by the dynamic time warping algorithm is a cumulative value compared with the point-to-point comparison of the local property of the signature according to the signature input time, unlike the distance between feature information determined as one value in the feature-based signature verification. The deviation of the distance value can be large. Therefore, it is necessary to find a relative distance for each user. To this end, the function-based signature verification module 180 calculates the distance (D (m) _reference ) between the reference names in the same manner as in Equation 4, and divides the distance between the reference name and the input signature by the distance between the reference names and normalizes it. do.

Here, D (m) _reference is the m-th feature information included in the feature information subset, and means the maximum value of the values obtained by the dynamic time warping algorithm for each feature information. DTW (R _{i , m} , R _{j , m} ) calculates the distance between the m th characteristic information R _{i , m} of the i th reference signature and the m th characteristic information R _{j , m} of the j th reference signature through a dynamic time warping algorithm. The value obtained. At this time, i and j are not the same. The reason for dividing by the maximum value of the distance between the reference signatures is to normalize the cumulative distance as a ratio to the maximum allowable range of error between the reference signatures.

Referring back to FIG. 2, the hybrid online signature verification system according to an embodiment of the present invention performs verification of an input signature through serial fusion of a feature-based signature verification technique and a function-based signature verification technique in the signature verification result fusion module 190. To perform.

Specifically, in the serial fusion method, the distance (D _feature ) between the reference name and the input signature for the feature-based signature verification in the first verification step (S202) is referred to as a tolerance criterion (θ _feature ) (hereinafter referred to as 'feature-based criterion'). If it is larger than the feature-based criterion (θ _feature ), the distance from the registered reference signature is out of tolerance. On the other hand, when the distance (D _feature ) between the reference name and the input signature is within the _feature- based criterion (θ _feature ), the distance between the reference name and the input signature (D) as a function-based signature verification method in the second verification step (S203). _function ) is compared with an acceptance criterion (θ _function ) (hereinafter referred to as a “function based criterion”). If the distance between the reference signature and the input signature (D _function ) is within the function-based criterion (θ _function ), it is authenticated with the true signature and in other cases it is determined as a false signature.

One embodiment of the present invention uses a serial fusion method in which a feature-based signature verification technique and a function-based signature verification technique are serially connected. In general, the feature-based signature verification technique is faster than the function-based signature verification technique. The overall form and characteristics of the signature make it possible to quickly determine the input signature, which is clearly different from the registered reference signature of the user, as a fake signature. That is, in the function-based signature verification, the cumulative distance representing the difference between two signatures is calculated by accumulating the difference in feature comparison according to the signature input time to calculate the distance between the user's registered reference name and the input signature. Verification takes longer than feature-based signature verification, which compares features and expresses them as distances. Therefore, the serial fusion method according to an embodiment of the present invention takes a function that takes a long time to verify signature in the second verification step (S203) only for the input signature that successfully passed the feature-based signature verification, which is the first verification step (S202). Because it performs the based signature verification, it has the advantage of quickly detecting the forged signatures only by the first step verification of serial fusion.

Hybrid on-line signature verification system according to an embodiment of the present invention is input based on the feature-based reference value (θ _feature ) and the function-based reference value (θ _function ) used in the feature-based signature verification method and function-based signature verification method You can control the degree to which the input signature is determined to be a true signature or a false signature. How to determine the feature-based reference value (θ _feature ) and the function-based reference value (θ _function ) can be an important parameter that affects the accuracy of the online signature verification system.

One embodiment of the present invention uses data mining techniques to establish the reliability of signature verification.

 In order to apply the data mining technology, the real name is collected from the user who uses the online signature verification, and the real name is collected and the real name is used as the learning data. At this time, the number of the true signature and the name of the go may be different.

FIG. 3 is an example of learning data, in which 30 authentic characters are collected for each of 70 users (u1 to u70), and 30 garrison names are elaborately copied from each true signature. In the training data collected for the actual online signature verification, select k (k = 5) reference character names (301) to be used as reference names among each user's (u1) certificate names, and the other character signatures (25) ) Is used as the learning signature 302. In addition, 30 go-names imitating the authenticity of the user u1 are used as the professional forgery signature 303, and randomly 30 of the authenticity signatures of the other users u2 to u70 are selected. Take advantage of. In this case, the forgery signature 303 and the random forgery signature 304 are used as a learning name, using only the forgery signature 303 as the learning name, or if necessary, using the signature forgery 303 and randomly. The fake signature 304 can be arbitrarily mixed and used as a learning name. In the following description, a learning goggle name in which a forgery signature 303 and a random forgery signature 304 are mixed is referred to as a 'full text + random forgery signature'.

4 is a view for explaining the entire process of the FAR and FRR setting method using a data mining technique. Here, the distance calculation method between the signatures on the training data may use the above-described D _feature and D _function calculation methods according to a signature verification method, that is, a feature-based signature verification method or a function-based signature verification method.

First, the maximum value is defined as max_D _r by calculating the distance D _r between the reference document names selected for each user (S401). Calculate the distance (D _g ) between the reference names and the learning name (S402) and define the minimum value as min_D _g and calculate the distance (D _f ) between the reference names and the study names (S403). ) The minimum value is defined as min_D _f .

을 계산하여 이를 해당 사용자의 '최소 진서명 거리비율'이라고 정의하고 모든 사용자의 최소 진서명 거리비율을 오름차순으로 정렬한다(S404). 각 사용자 별로

을 계산하여 이를 해당 사용자의 '최소 가서명 거리비율'이라고 정의하고 모든 사용자의 최소 가서명 거리비율을 오름차순으로 정렬한다(S405).For each user

By calculating this, it is defined as the 'minimal signature distance ratio' of the user and sorts the minimum true signature distance ratio of all users in ascending order (S404). For each user

By calculating this, it is defined as the 'minimum go street distance ratio' of the user and sorts the minimum go street distance ratio of all users in ascending order (S405).

In addition, when a threshold value is arbitrarily raised from 0 to a list of the minimum true signature distance ratios of all users, the minimum true signature distance ratio is larger than the corresponding threshold and thus cannot be recognized as a true signature (ie, a false reject error. The ratio of the user is calculated and defined as a false rejection ratio (FRR) at the corresponding threshold (S406). In the same way, when a random threshold value is raised from zero to the minimum garrison distance ratio list of all users in order, the minimum garrison distance ratio is smaller than the threshold and thus is falsely recognized as a true signature (i.e., false signature authentication error) acceptance error)) The ratio of the user is calculated and defined as a FAR (false acceptance ratio) at the threshold (S407).

Accordingly, an embodiment of the present invention may utilize FRR and FAR calculated using data mining techniques as feature-based reference values (θ _feature ) and function-based reference values (θ _function ) to set the reliability of signature verification. .

Applications that require on-line signature verification may use different signature verification reliability depending on the content and importance of the application service. For example, if you authenticate yourself to provide important services, you should use a high degree of signature verification. If you provide non-critical services, signature verification is either formal or does not perform signature verification. You may be satisfied with storing. Also, different signature verification reliability can be used depending on the complexity of the user's signature. For example, signature verification may be performed with low reliability for users who use complex signatures that are difficult to forge, and signature verification may be performed with high reliability for users who use simple signatures that are easy to forge. Therefore, in one embodiment of the present invention, in the on-line hybrid signature verification system, the feature-based signature verification technique and the function-based signature verification technique are serially fused and the reference values θ _feature and θ _function used in each scheme are allowed to be applied. Depending on the complexity or complexity of the user registration signature, signature verification can be performed based on different reliability criteria.

For example, in one embodiment of the present invention, a total of eight reliability criteria can be set from 0 to 7 security levels as shown in Table 1 as a systematic multi-standard signature verification reliability to support signature verification of various applications. The multi-criterion signature verification reliability is not set arbitrarily, but is a reliability reference value systematically set by using the method described with reference to FIGS. 3 and 4 by using a data mining technique for an online signature verification data set.

Security level θ _feature θ _function Setting criteria Ⓞ ∞ ∞ Do not run signature verification ① 1.192 ∞ Using only the signature-based signature verification technique, using the full + random counterfeit signature as a learning name, set the minimum distance value of FRR = 0 to θ _feature to perform only the first step of serial fusion. ② 1.192 1.024 Use the θ _feature value of ① as the criterion of the first stage of serial fusion, and use the distance value of EER using the full + random counterfeit signature as a learning name for the function-based reliability criterion of the second stage as the θ _function . ③ 1.192 0.967 Using the θ _feature value of ① as the criterion of the first stage of serial fusion, and using the distance value of EER using only professional forgery signature as the learning name for the function-based reliability criterion value of the second stage, as the θ _function . ④ 1.192 0.866 Using the θ _feature value of ① as the criterion of the first stage of serial fusion, and using only the professional forgery signature as a learning name for the function-based reliability criterion value of the second stage, the minimum distance value of FAR = 7 is used as the θ _function . use ⑤ 1.192 0.697 Using the θ _feature value of ① as the criterion of the first stage of serial fusion, and using only the professional forgery signature as a learning name for the function-based reliability criterion value of the second stage, the minimum distance value of FAR = 2 is used as the θ _function . use ⑥ 1.192 0.637 Using the θ _feature value of ① as the criterion of the first stage of serial fusion, and using only the professional counterfeit signature as a learning name for the function-based reliability criterion value of the second stage, the minimum distance value of FAR = 1 as θ _function use ⑦ 1.192 0.510 Using the θ _feature value of ① as the criterion of the first stage of serial fusion, and using only the professional counterfeit signature as a learning name for the function-based reliability criterion value of the second stage, the minimum distance value of FAR = 0 is used as the θ _function . use

In security level 0, authentication is performed with a true signature without performing the actual verification process.In security level 1, only the feature-based signature verification step, which is the first step of serial convergence, is performed. In the Level 2 level, the reliability criteria for later feature-based signature verification are set to be the same as the Security Level 1 level, and multi criteria are set by gradually increasing the signature verification reliability level of the second function-based signature verification level to the Security Level 7 level. It is. In Table 1, FAR means false acceptance ratio and FRR means false rejection ratio. EER (equal error ratio) means a point where the ratio of false signature authentication errors and the true signature rejection errors are the same.

The reasons for selecting each security level and the applicable fields of the established multi-criterion signature verification reliability are as follows.

Security level ⓞ: All the input signature for verification entered for verification does not carry out signature verification. In other words, in the serial fusion, the security level is interpreted as having passed the first and second steps unconditionally, so that all input verification input signatures are artificially determined as true signatures, which does not require actual signature verification for security verification. It can be used in applications that simply store the signed signature.

Security level ①: It is a security level that is determined as a true signature if only the first verification process is passed without the second verification process in serial fusion. It is a security level that recognizes as a true signature if the overall form and characteristics of the signature are similar. That is, in the feature-based signature verification technique, which is the first step of serial fusion, the main purpose is to quickly reject the counterfeit signature. Therefore, in order to prevent possible false reject error, the minimum distance value of FRR = 0 is calculated. _feature It is used as a value (see FIG. 5). This security level can be used in applications that quickly determine the authenticity of online signatures using only feature-based signature verification techniques that compare similarities only with the overall form and global characteristics of the signature.

Security level ②: The lowest security level among the security levels that use both the first and second verification processes of serial convergence. Θ _feature of security level ① Feature-based signature verification is performed by value, and function-based signature verification is additionally performed on the signature that passes. Θ _function based on the function signature-based verification techniques The value is set to the distance value of the point (EER) where the FRR and the FAR obtained by using a 'special + random counterfeit signature', which arbitrarily mixes a forgery signature and a random forgery signature, as the learning names (see FIG. 6). This security level is used in applications that attempt to determine the authenticity of the input signature more accurately by comparing features over time for signatures that pass the feature-based signature verification technique that compares similarities only with the overall form and global features of the signature. Can be used.

Security level ③: Based on the security level ② similar but function-based signature verification techniques θ _function This is a learning gag name for setting the value, and using only the forgery signature, it is set to the distance value of the point (EER) where FRR and FAR are equal. Since only the forgery signature is included in the learning name, the FAR is higher than the case where the random forgery signature is included, thereby reducing the ERR point (see FIG. 6). That is, it can be used in applications that need to perform the second verification process of serial fusion through more accurate function-based signature verification.

Security level ④: ③ security level is similar but based on function-based signature verification techniques θ _function It is set as the distance value of the point where the FAR is 7% using only the fake signature as a learning name for setting the value (see FIG. 6). The reason for setting the FAR to 7% is based on various developments to date for the signature verification data set (coordinate value (x, y) and pressure value p according to signature time t) collected as described in FIG. When classified using the data mining technique, experimental results show that the accuracy is up to 93%. Therefore, the tolerance level is set to 7%. This security level is appropriate when the signature input device used for signature verification provides coordinate and pressure values over the signature time.

Security level ⑤: Similar to security level ④, but based on function-based signature verification technique θ _function It is set as the distance value of the point where the FAR is 2% using only a fake signature as a learning name for setting the value (see FIG. 6). The rationale for setting FAR to 2% is the ideal signature verification data set used to compare the performance of the signature verification algorithm (coordinate values (x, y) over signature time t, pressure value p, pen angle, and pen tilt). When the data is classified using various data mining techniques developed to date, the results show that the accuracy is up to 98%. Therefore, the tolerance level is set to 2%. This class can be used as a signature verification reliability for applications that use signature input devices that can provide the angle and tilt of the pen in addition to coordinate and pressure values over signature time.

Security level ⑥: Similar to security level ⑤, but based on function-based signature verification technique θ _function It is set as the distance value of the point where FAR is 1% using only a fake signature as a learning name for setting a value (see FIG. 6). The reason for setting FAR to 1% is about 99% accuracy target of signature verification algorithms studied so far. Therefore, the tolerance level is set to 1%. This class is the reliability figure targeted by current signature verification techniques and can be used in applications that require high signature verification reliability.

Security level ⑦: Similar to security level ⑥, but based on function-based signature verification technique θ _function It is set as the distance value of the point where FAR is 0% using only a fake signature as a learning name for setting a value (see FIG. 6). Since 0% of FAR means that false signatures are incorrectly classified and authenticated as true signatures, it is not a probability. Therefore, the feature-based signature verification is based on the point where FRR = 0. In the verification, it can be said that the signature verification criteria with perfect accuracy by using the point with FAR = 0. Thus, in this class, the signature verification reliability is not easily passed by the user who is inconsistent with the signature. In other words, it provides a very important service can be used in applications where user identification is important.

7 illustrates a parallel fusion process of outputting a similarity value (0 to 100) between a user's reference name and a verification input signature by fusing a distance value by a feature-based signature verification technique and a distance value by a function-based signature verification technique. will be.

First, a distance (D _feature , D _function ) between the reference name and the input name according to the feature-based signature verification technique and the function-based signature verification technique is calculated (S701). The process of calculating the distance (D _feature , D _function ) between the reference name and the input name is the same as the calculation method according to Equations 1 to 4 described with reference to FIG. 2.

And, for the feature-based signature verification, FAR corresponding to the distance between the reference name and the input name (D _feature ) Extract the _feature value (S702) and FAR corresponding to the distance between the reference name and the input name (D _function ) for function-based signature verification _{The function} value is extracted (S703).

Verification result of feature-based signature verification technique (FAR) _feature ) and verification results of function-based signature verification (FAR) _A similarity between the reference signature name and the input signature name for determining the signature authenticity is calculated using a _function ) (S704). In this case, the similarity between the reference name and the input name may be defined as shown in Equation 5.

One embodiment of the present invention integrates the FAR _{feature of the feature-} based signature verification technique and the FAR _function of the function-based signature verification technique and presents them to the user as one similarity value. Is a fusion method that can be used as a complementary method, and is suitable for the application that wants to express the authenticity of the signature in a fuzzy step by seeing how similar the input signature is to the registered reference signature of the user.

8 and 9 illustrate an example of a user interface for a multi-criteria reliability support hybrid online signature verification system according to an embodiment of the present invention.

Interface to output the authenticity of the signature according to the reliability according to the characteristics and service importance of the application requiring the online signature verification, or the complexity and repetitive consistency of the user signature (FIG. 8) or the input signature and the registered reference signature It is possible to configure the interface (Fig. 9) for outputting a similarity value with. In the case of multi-criterion signature verification reliability, the reliability button 801 is selected. In order to make fuzzy decisions based on the similarity, the similarity button 901 is selected. These button selections can be selected manually by the user, or can be set automatically by the system after the situation has been determined. In addition, when multi-criteria reliability is selected (FIG. 8), an administrator may set an appropriate security level, and the system may dynamically select a security level according to the environment. When similarity is selected (FIG. 9), the authenticity of the input signature can be determined fuzzy according to the output similarity value.

One embodiment of the present invention can support multi-criteria reliability by fusing a specific-based signature verification technique and a function-based signature verification technique, and enable on-line signature verification by dynamically adjusting the reliability of signature verification by the system automatically or user-selectively. It can support the accuracy and security of signature verification suitable for the application environment.

Embodiments of the present invention include computer readable media including program instructions for performing various computer implemented operations. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The media may be program instructions that are specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

As described above, the present invention has been described by specific embodiments such as specific components and the like. For those skilled in the art to which the present invention pertains, various modifications and variations are possible.

Therefore, the spirit of the present invention should not be limited to the described embodiments, and all of the equivalents or equivalents of the claims as well as the claims to be described later will belong to the scope of the present invention. .

1 illustrates an internal configuration of a hybrid online signature verification system according to an embodiment of the present invention.

2 is a diagram illustrating a signature verification process through serial fusion in a hybrid online signature verification system according to an embodiment of the present invention.

FIG. 3 is a diagram for describing learning data used to set multi-criteria reliability. FIG.

4 is a diagram for describing a method of setting multi-criteria reliability using data mining technology.

5 and 6 are diagrams for explaining the multi-criterion signature verification reliability setting process according to the security level.

7 is a diagram illustrating a signature verification process through parallel fusion in a hybrid online signature verification system according to an embodiment of the present invention.

8 and 9 illustrate an example of a user interface for a multi-criteria reliability support hybrid online signature verification system according to an embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

170: feature-based signature verification module

180: function based signature verification module

190: Signature verification result fusion module

Claims (9)

delete delete A feature-based signature verification module for calculating a distance calculation value of the feature-based signature verification, which is a distance between the vector extracted from the reference signature and the vector extracted from the input signature; A function-based signature verification module for calculating a distance calculation value of a function-based signature verification, which is a distance obtained by accumulating the difference between the reference signature and the input signature according to a signature input time at which the reference signature and the input signature are input; And, A signature verification result fusion module for determining the authenticity of the input signature using the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification. Including, The signature verification result fusion module, Acceptance threshold value, which is a criterion for calculating the false rejection ratio (FRR) and false acceptance ratio (FAR) for the reference signature using data mining, and for determining the authenticity of the input signature according to the calculated FRR and FAR. Multi-criteria reliability support hybrid online signature verification system to set up. A feature-based signature verification module for calculating a distance calculation value of the feature-based signature verification, which is a distance between the vector extracted from the reference signature and the vector extracted from the input signature; A function-based signature verification module for calculating a distance calculation value of a function-based signature verification, which is a distance obtained by accumulating the difference between the reference signature and the input signature according to a signature input time at which the reference signature and the input signature are input; And, A signature verification result fusion module for determining the authenticity of the input signature using the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification. Including, The signature verification result fusion module, The similarity of the reference signature and the input signature is calculated using the FAR corresponding to the distance calculation value of the feature-based signature verification and the FAR corresponding to the distance calculation value of the function-based signature verification, and the similarity of the input signature is calculated according to the calculated similarity. Multi-criteria reliability support hybrid online signature verification system for determining authenticity. delete delete In the multi-line reliability support hybrid online signature verification method of a hybrid online signature verification system including a feature-based signature verification module, a function-based signature verification module, and a signature verification result fusion module, Calculating, by the feature-based signature verification module, a distance calculation value of feature-based signature verification, which is a distance between a vector extracted from a reference signature and a vector extracted from an input signature; Calculating, by the function-based signature verification module, a distance calculation value of a function-based signature verification, which is a distance obtained by accumulating the difference between the reference signature and the input signature according to the signature input time at which the reference signature and the input signature are input; And, Determining the authenticity of the input signature by using the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification in the signature verification result fusion module. Including, Determining the authenticity of the input signature, Calculating FRR and FAR for the reference signature using data mining; Setting an acceptance criterion value as a criterion for determining the authenticity of the input signature according to the calculated FRR and FAR; Determining the authenticity of the input signature by comparing the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification with the set acceptance criteria. Including, multi-criteria reliability support hybrid online signature verification method. The method of claim 7, wherein Computing the FRR and the FAR for the reference signature, Calculating a maximum true signature distance, the maximum distance among the distances between the reference signatures; Calculating a minimum true signature distance that is a minimum distance among the distances between the reference signatures; Calculating a minimum garish name distance, which is a minimum distance between distances between the reference name and the name of the forged name; Calculating a minimum true signature distance ratio, which is a ratio of the maximum true signature distance and the minimum true signature distance; Calculating a minimum garrison distance ratio, which is a ratio of the maximum true signature distance and the minimum garrison distance, Calculating an FRR using the minimum signature distance ratio; Calculating a FAR using the minimum go-go distance ratio Including, multi-criteria reliability support hybrid online signature verification method. In the multi-line reliability support hybrid online signature verification method of a hybrid online signature verification system including a feature-based signature verification module, a function-based signature verification module, and a signature verification result fusion module, Calculating, by the feature-based signature verification module, a distance calculation value of feature-based signature verification, which is a distance between a vector extracted from a reference signature and a vector extracted from an input signature; Calculating, by the function-based signature verification module, a distance calculation value of a function-based signature verification, which is a distance obtained by accumulating the difference between the reference signature and the input signature according to the signature input time at which the reference signature and the input signature are input; And, Determining the authenticity of the input signature by using the distance calculation value of the feature-based signature verification and the distance calculation value of the function-based signature verification in the signature verification result fusion module. Including, Determining the authenticity of the input signature, Extracting a FAR corresponding to the distance calculation value of the feature-based signature verification; Extracting a FAR corresponding to the distance calculation value of the function-based signature verification; Calculating the similarity between the reference name and the input signature using the FAR corresponding to the distance calculation value of the feature-based signature verification and the FAR corresponding to the distance calculation value of the function-based signature verification; Determining the authenticity of the input signature according to the calculated similarity Including, multi-criteria reliability support hybrid online signature verification method.

Images