JP4283864B1 - User authentication program and user authentication system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a user authentication program and a user authentication system capable of continuously executing user authentication processing by monitoring key operation information and reducing a probability that a registered user is erroneously rejected from authentication.
A user determination data storage unit 42 that stores user determination data having a triplet and a first ratio value, a key operation information acquisition unit 51 that acquires key operation information, and a triplet extraction unit that extracts a triplet. 52, a time data acquisition unit 53 that acquires time data, a ratio value calculation unit 55 that calculates a first ratio value, a triplet data registration unit 57 that registers triplet data in the triplet data storage unit 43, and determination target data A determination target data acquisition unit 64 that acquires the evaluation value, an evaluation value calculation unit 65 that calculates an evaluation value, a feature value calculation unit 66 that calculates a feature value of the determination target data, and a user determination unit 67 that determines the determination target data Make the computer work.
[Selection] Figure 1

Description

  The present invention relates to an authentication technique for a user who uses a computer, and more particularly, to a user authentication program and a user authentication system for monitoring operation information from a keyboard and continuously executing a user authentication process.

  Conventionally, as a user authentication technique, a technique for performing login authentication by a combination of a user ID and a password is known. However, in this login authentication technique, there is a problem that if a password is leaked or predicted by another person, it can be easily broken. In addition, since authentication is performed only at the start of use of the computer, there is a problem that anyone can easily operate once a user-authenticated system is left.

  In particular, it has been pointed out that the above-mentioned login authentication alone is not sufficient in a system that handles document data that cannot be “spoofed” or tampered by others, such as an electronic medical chart that has been attracting attention in recent years. Because in hospital systems, it is not realistic for doctors and nurses who have received emergency calls from patients to perform logout processing thoroughly, and there are many systems that remain logged in. is there.

  In order to solve such a problem of the login authentication technique, a technique (keystroke dynamics technique) for sequentially authenticating a user based on a bag or a feature included in input operation information from a keyboard has been proposed. For example, in Japanese Patent Laid-Open No. 2003-58504, user operation information in an input unit for an information processing apparatus is acquired, personal operation feature information is extracted from the operation information, the extracted individual operation feature information, and storage Permission information for executing collation processing with the operation feature information of the registered user stored in the means, and determining whether or not to execute the process in the information processing apparatus corresponding to the user operation on the input means based on the result of the collation processing And a personal authentication processing device having a configuration for outputting the generated permission information to the information processing device is disclosed (Patent Document 1).

JP 2003-58504 A

  However, in the conventional keystroke dynamics technology including the invention described in Patent Document 1, there is a high probability that a registered user is erroneously rejected for authentication, and the usability is poor. Specifically, since the input operation information is used for collation processing without any selection, the input rhythm is delayed by pressing the shift key or space key (conversion key), or the input operation is accidentally interrupted. In the case of a character string that can be input at high speed by frequent input operations such as the user's name or the like, authentication is rejected even though the user is a registered user.

  In Patent Document 1, as operation information for user verification, the time from pressing the (n-1) th key to the nth key (DDn), the time from pressing the nth key to releasing (UDn), Alternatively, the average value of the time (DUn) from the n−1th key release to the nth key press is simply used. For this reason, there is a problem that the collation result is very easily affected by the variation in the type speed, and even if the input speed is slightly different, there is a high possibility that even a registered user will be rejected.

  Furthermore, in Patent Document 1, the input operation information is directly registered in the authentication information database. For this reason, if operation information is continuously input, there is a possibility that a high load is applied to the computer and the network. In addition, since the input character strings are registered in the order as they are, if the data is stolen, there is a risk that the input content may be easily analyzed, and there is also a security concern.

  The present invention has been made in order to solve such problems. In addition to monitoring operation information from a keyboard and continuously executing user authentication processing, a registered user erroneously authenticates. An object of the present invention is to provide a user authentication program and a user authentication system that can reduce the probability of being rejected.

A feature of the user authentication program according to the present invention is that a triplet consisting of a combination of three consecutive press keys acquired from the operation information of the keyboard operated by the registered user and the first key of the triplet operation information is pressed. The first time calculated as the ratio of the first waiting time (T ₁ ) until the second key is pressed and the second waiting time (T ₂ ) from when the second key is pressed until the third key is pressed. User determination data storage unit that stores in advance user determination data having at least a ratio value (R ₁ ) in association with the registered user, and key operation information acquisition for acquiring key operation information when the keyboard is operated A triplet extracting unit that extracts a predetermined triplet from the key operation information, and the first triplet extracted by the triplet extracting unit. And Chi time (T ₁₎ and the second waiting time (T ₂₎ time data acquisition unit that acquires a ratio value calculation unit for calculating a first ratio value (R ₁₎ for each triplet, each triplet A triplet data registration unit that registers the triplet data in the triplet data storage unit as triplet data associated with the first ratio value (R ₁ ), and a determination that acquires the triplet data registered in the triplet data storage unit as determination target data Evaluation that calculates an evaluation value (V) of each first ratio value (R ₁ ) in the determination target data based on the target data acquisition unit and the user determination data stored in the user determination data storage unit Based on the value calculation unit and each evaluation value (V) calculated by the evaluation value calculation unit for all triplets included in the determination target data The determination target data is input by the registered user based on a feature value calculation unit that calculates a feature value (C) related to the determination target data, and a magnitude relationship between the feature value (C) and a predetermined cutoff value. The point is to make the computer function as a user determination unit for determining whether or not it has been performed.

In the present invention, in the user determination data storage unit, a second ratio value (R ₂ ) obtained as a ratio of pressing times of any two pressing keys among three consecutive pressing keys constituting the triplet is stored. Further stored as user determination data, the time data acquisition unit acquires the pressing time (D ₁ , D ₂ , D ₃ ) of each key for each triplet extracted by the triplet extraction unit, and the ratio The value calculation unit calculates the second ratio value (R ₂ ) for each triplet, and the triplet data registration unit associates each triplet with the second ratio value (R ₂ ) and stores the triplet data as triplet data. The evaluation calculation unit is configured to store the determination target data based on the user determination data stored in the user determination data storage unit. The second ratio value of the inner calculates an evaluation value (V) for _{(R 2),} wherein value calculation unit, an evaluation value for said first ratio value _{(R 1)} and a second ratio value _{(R 2)} (V ), It is preferable to make the computer function so as to calculate the characteristic value (C) regarding the determination target data.

Furthermore, in the present invention, the first waiting time (T ₁ ) and the second waiting time (T ₂ ) or the pressing time of each key (D ₁ , D ₂ , D) in each triplet acquired by the time data acquisition unit. D ₃ ) includes a triplet sorting unit that sorts triplets within a predetermined time limit. For each triplet sorted by the triplet sorting unit, the ratio value calculation unit uses the first ratio value (R ₁ ) and the second ratio value (R ₂ ) are calculated, and the triplet data registering unit associates the two and registers them as triplet data in the triplet data storage unit.

  In the present invention, the evaluation value calculation unit compares the triplet ratio value included in the determination target data with the ratio value of the same triplet group stored in the user determination data. The computer is calculated so that a higher evaluation value (V) is calculated as the ranking is closer to the median value, and a lower evaluation value (V) is calculated as the ranking is closer to the highest value or the lowest value. It is preferable to make it function.

Furthermore, in this invention, it is preferable that the said evaluation value calculation part makes a computer function so that the said evaluation value (V) may be calculated by following formula (1).
V (R) = a−b × | (LS) | × (L + S) Expression (1)
However,
a, b: Coefficient R: First ratio value or second ratio value L: Among triplets in the user determination data, the same character string as the triplet of the determination target data, and a ratio value larger than the ratio value of the triplet S: Number of triplets in the user determination data having the same character string as the determination target data triplet and having a ratio value less than or equal to the ratio value of the triplet

In the present invention, it is preferable that the feature value calculation unit causes the computer to function so as to calculate the feature value (C) by the following equation (2).
C = {Σ [W × V (R)]} / N (2)
However,
W: Weight value determined based on the number of triplets of the same character string as the triplet of the determination target data among the triplets in the user determination data N: Number of triplets of the unit to be evaluated

  Furthermore, in the present invention, the time data acquisition unit further acquires, for each triplet extracted by the triplet extraction unit, a total triplet time from when the first key is pressed until the third key is released. The triplet sorting unit preferably causes the computer to function so as to sort only triplets within a predetermined time limit from the acquired total triplet time.

  Further, in the present invention, before registering the triplet data of the triplets extracted by the triplet extraction unit in the triplet data storage unit, the computer is used as an extraction order randomizing unit that randomly rearranges the extraction order for each predetermined number of triplet groups. It is preferable to make it function.

Furthermore, in the present invention, a basic data acquisition unit that acquires basic data for creating user determination data related to a predetermined registered user from the triplet data stored in the triplet data storage unit, and included in the basic data A triplet number calculating unit for calculating the number of triplets of the same character string, a weight value setting unit for setting a weight value for each triplet according to the number of triplets calculated by the triplet number calculating unit, and for each triplet In addition, a variation coefficient calculation unit that calculates a variation coefficient (standard deviation / average value) of the first ratio value (R ₁ ) to the second ratio value (R ₂ ), and a weight value set by the weight value setting unit Is a predetermined value or more, and only the triplet data whose variation coefficient calculated by the variation coefficient calculation unit is equal to or smaller than the predetermined value is used as the user determination data. It is preferable to make the computer function as a triplet re-sorting unit that re-sorts as a data.

In addition, the user authentication system according to the present invention includes a triplet composed of a combination of three consecutive press keys acquired from the operation information of the keyboard operated by the registered user, and the first key of the triplet operation information is pressed. The first time calculated as the ratio of the first waiting time (T ₁ ) until the second key is pressed and the second waiting time (T ₂ ) from when the second key is pressed until the third key is pressed. User determination data storage unit that stores in advance user determination data having at least a ratio value (R ₁ ) in association with the registered user, and key operation information acquisition for acquiring key operation information when the keyboard is operated A triplet extracting unit that extracts a predetermined triplet from the key operation information, and each triplet extracted by the triplet extracting unit. Time (T ₁₎ and the second waiting time (T ₂₎ time data acquisition unit that acquires a ratio value calculation unit for calculating a first ratio value (R ₁₎ for each triplet, the each triplet A triplet data registration unit that registers in the triplet data storage unit as triplet data associated with the first ratio value (R ₁ ), and a determination target that acquires the triplet data registered in the triplet data storage unit as determination target data An evaluation value for calculating an evaluation value (V) of each first ratio value (R ₁ ) in the determination target data based on the data acquisition unit and the user determination data stored in the user determination data storage unit Based on each evaluation value (V) calculated by the evaluation value calculation unit for all triplets included in the calculation unit and the determination target data The determination target data is input by the registered user based on a feature value calculation unit that calculates a feature value (C) related to the determination target data, and a magnitude relationship between the feature value (C) and a predetermined cutoff value. And a user determination unit that determines whether or not the device is a device.

  According to the present invention, the operation information from the keyboard can be monitored and the user authentication process can be continuously executed, and the probability that the registered user is mistakenly rejected for authentication can be reduced.

  Embodiments of a user authentication program and a user authentication system according to the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing the overall configuration of a user authentication system 1 equipped with a user authentication program 1a according to the present invention. As shown in FIG. 1, a user authentication system 1 according to the present embodiment mainly includes an input unit 2 including a keyboard and a mouse, a display unit 3 including a liquid crystal display, a user authentication program 1a according to the present embodiment, and various types. Storage means 4 for storing the above-described data and the like, and arithmetic processing means 5 for controlling each of these constituent means and acquiring various data and executing arithmetic processing.

  The storage means 4 is composed of a hard disk, a RAM (Random Access Memory), etc., and has a program storage unit 41, a user determination data storage unit 42, and a triplet data storage unit 43 as shown in FIG. is doing.

  Hereinafter, each component of the storage unit 4 will be described in more detail. The user authentication program 1a of this embodiment is installed in the program storage unit 41. And it is made to function as the user authentication system 1 of this embodiment by being performed by the arithmetic processing means 5. In the present embodiment, all functions to be described later are realized by the user authentication program 1a. However, the present invention is not limited to this, and each function may be made independent as a separate program.

  The user determination data storage unit 42 stores in advance user determination data for determining whether or not the key operation from the keyboard is derived from the registered user. The user determination data indicates the typing habits and characteristics of the registered user, and is created based on triplet data sampled from the registered user's key operation information, as will be described later.

  Further, a combination of three consecutive pressed keys (referred to as “triplet” in the present invention) is adopted as a data unit constituting the user determination data. This is because, as a result of numerous sampling investigations by the inventor of the present application, when a combination of two consecutive press keys is used, the parameter is not stable and the probability of erroneously refusing a registered user is high. On the other hand, when the combination of four consecutive press keys is used, it has been found that the data becomes enormous and it takes too much time to create and determine the data. Therefore, the inventors have found that it is preferable to perform authentication determination using a combination of three consecutive pressed keys.

  Further, in the present embodiment, the triplet is a combination of three consecutive press keys consisting of only 26 types of lowercase alphabet keys (az). Keys other than alphabet keys, such as shift keys, space keys, return keys, number keys, symbol keys, etc., may deviate from the home position in touch typing or operate the Kana-Kanji conversion system. This is because it is considered inappropriate as data indicating typing habits and features unique to registered users.

FIG. 2 is a diagram illustrating an example of user determination data in the present embodiment. In this embodiment, the user determination data is associated with a registered user to be authenticated, a triplet composed of a combination of three consecutive press keys (first key, second key, and third key), and the triplet Of the operation information, the first waiting time (T ₁ ) from when the first key is pressed to when the second key is pressed and the second time from when the second key is pressed until the third key is pressed The first ratio value (R ₁ ) obtained as the ratio of the waiting time (T ₂ ), and the _second ratio obtained as the ratio of the pressing time (D ₂ ) of the _second key and the pressing time (D ₃ ) of the third key. Value (R ₂ ).

In the present embodiment, the user determination data includes both the first ratio value (R ₁ ) and the second ratio value (R ₂ ), but the present invention is not limited to this. If it has 1 ratio value (R ₁ ), a predetermined effect can be obtained. As the second ratio value (R ₂ ), the ratio of the second key pressing time (D ₂ ) to the third key pressing time (D ₃ ) is used. A ratio of D ₁ ) and the second key pressing time (D ₂ ) or a ratio of the first key pressing time (D ₁ ) and the third key pressing time (D ₃ ) may be used. Furthermore, in this embodiment, as will be described later, triplets whose number of inputs is 10 or less are excluded from the viewpoint of ensuring the reliability of the sample.

The triplet data storage unit 43 stores triplet data sampled from the key operation information of the registered user. This triplet data is used as basic data for creating user determination data. In the present embodiment, the triplet data is also used as determination target data for determining whether or not the user is a registered user, as will be described later. Therefore, the triplet data is composed of a triplet, a first ratio value (R ₁ ), and a second ratio value (R ₂ ), as with the user determination data described above.

  Next, the arithmetic processing means 5 is composed of a CPU (Central Processing Unit) and the like, and mainly collects triplet data from key operation information by executing the user authentication program 1a installed in the program storage unit 41. A triplet data collection function for performing the user determination, a user determination data generation function for generating user determination data, and a user determination function for determining the user who operates the key are realized.

Hereinafter, each component for realizing each function will be described in more detail. First, as shown in FIG. 1, the triplet data collection function mainly includes a key operation information acquisition unit 51, a triplet extraction unit 52, a time data acquisition unit 53, a triplet selection unit 54, and a ratio value calculation unit 55. , an extraction sequence run Damaizu unit 56 are realized by causing a computer to function as triplet data registration unit 57.

  Hereinafter, each component that realizes the triplet data collection function will be described in more detail.

  The key operation information acquisition unit 51 acquires key operation information when the keyboard is operated. In this embodiment, the user authentication system 1 is configured to create internal message data each time a key on the keyboard is pressed or released. Therefore, the key operation information acquisition unit 51 hooks these message data and acquires them as key operation information.

  The triplet extraction unit 52 extracts a predetermined triplet from the key operation information. Specifically, the triplet extraction unit 52 extracts a triplet in which all of three consecutive pressed keys are composed of valid key groups from the key operation information acquired by the key operation information acquisition unit 51. . In the present embodiment, only 26 types of lowercase alphabet keys (az) are set as the valid key group, and other keys are set as the invalid key group.

  For example, from the key operation information “Keystroke dynamics”, as shown in FIG. 3, “eys” “yst” “str” “tro” “rok” “oke” “dyn” “yna” “nam” “ami” Twelve triplets “mic” and “ics” are extracted. In this case, since the first “K” is a capital letter, the key operation information “Key” is not extracted as a triplet. In this example, key operation information “ke_”, “e_d”, and “_dy” including a space key (blank part) are not extracted as a triplet.

The time data acquisition unit 53 acquires predetermined time data for each pressed key constituting the triplet. Specifically, for each triplet extracted by the triplet extraction unit 52, the time data acquisition unit 53, as shown in FIG. 4, the first key pressing time (D ₁ ) and the second key pressing time (D ₂ ), the pressing time (D ₃ ) of the third key, the first waiting time (T ₁ ) from when the first key is pressed until the second key is pressed, and the second key is pressed And the second waiting time (T ₂ ) until the third key is pressed.

Further, in the present embodiment, the time data acquisition unit 53, for each triplet extracted by the triplet extraction unit 52, totals from when the first key is pressed until the third key is released, as shown in FIG. The triplet time (T _A ) is also acquired.

The triplet sorting unit 54 sorts only effective triplets by excluding exceptional triplets. Specifically, the triplet sorting unit 54 acquires each time data (D ₁ , D ₂ , D ₃ , T ₁ , T ₂ , T _A ) acquired by the time data acquiring unit 53, and each time data is predetermined. It is determined whether it is included within the time limit. Only triplets whose time data are within the time limit are selected as appropriate triplets. On the other hand, when any time data deviates from the time limit, the triplet is not adopted as exceptional data.

In the present embodiment, as the time limit, an upper limit value and a lower limit value for each key pressing time (D ₁ , D ₂ , D ₃ ), and an upper limit value and a lower limit value for each waiting time (T ₁ , T ₂ ). , And an upper limit value for the total triplet time (T _A ) from when the first key is pressed until the third key is released, but is not limited to this, and at least the first waiting time ( A time limit may be set for T ₁ ) and the second waiting time (T ₂ ) or each key press time (D ₁ , D ₂ , D ₃ ), and the triplets may be selected.

The ratio value calculation unit 55 calculates the above-described first ratio value (R ₁ ) and second ratio value (R ₂ ) for each triplet. Specifically, the ratio value calculation unit 55 calculates the first waiting time (T ₁ ) and the second waiting time (T ₂ ) acquired by the time data acquisition unit 53 for each triplet selected by the triplet selection unit 54. To obtain a first ratio value (R ₁ = T ₁ / T ₂ ). Similarly, the ratio value calculation unit 55 acquires any two of the pressing times of the key pressing times (D ₁ , D ₂ , D ₃ ) acquired by the time data acquisition unit 53 and obtains the second ratio value ( R ₂ = D ₂ / D ₃ , D ₁ / D ₂ , D ₁ / D ₃ ).

  In the present embodiment, the ratio value calculation unit 55 calculates, as an internal process, an integer obtained by multiplying one by 256 and then dividing by the other and rounding off the decimal part as the ratio value. . This speeds up the calculation.

Extraction sequence run Damaizu 56 for difficult to reproduce keypress string, it is intended to sort the extracted sequence of triplets randomly. In this embodiment, the extraction order randomizing unit 56 attaches a header including a time stamp (in seconds) to each triplet before registering the triplet data of each triplet selected by the triplet selection unit 54 in the triplet data storage unit 43. Then, sequentially, the data is temporarily stored in the RAM of the storage unit 4. Then, when a predetermined number of triplet groups are accumulated, the extraction order by the triplet extraction unit 52 is randomly rearranged.

In this embodiment, the number of triplets accumulated during randomization is set to 50 to 300 in consideration of the balance between the difficulty of reproduction and the processing speed, but may be increased or decreased as appropriate. The extraction sequence run Damaizu unit 56 are those provided Optionally, separately, it may be improved security the encryption processing program.

The triplet data registration unit 57 registers triplet data in the triplet data storage unit 43. Specifically, for each triplet selected by the triplet selection unit 54, the first ratio value (R ₁ ) and the second ratio value (R ₂ ) calculated by the ratio value calculation unit 55 are acquired, and these are associated with each other. Further, it is registered in the triplet data storage unit 43 as triplet data.

  In this embodiment, the triplet data registration unit 57, when registering triplet data as basic data for creating user determination data, has a sufficient number of triplets that can guarantee the reliability as indicating the user characteristics ( Secure about 700-10000) and register it in association with the user. Moreover, in this embodiment, the triplet data registration part 57 ensures the triplet number (about 50-400) suitable for one determination, when registering triplet data as determination object data at the time of user determination.

  Next, each component that implements the user determination data creation function will be described. As shown in FIG. 1, the user determination data creation function mainly includes a basic data acquisition unit 58, a triplet number calculation unit 59, a weight value setting unit 60, a variation coefficient calculation unit 61, and a triplet re-sorting unit. 62, and by causing the computer to function as the user determination data registration unit 63.

  Hereinafter, each component that implements the user determination data creation function will be described in more detail.

  The basic data acquisition unit 58 acquires basic data for creating user determination data. Specifically, the basic data acquisition unit 58 acquires, as basic data, triplet data collected for a predetermined registered user who wants to create user determination data among the triplet data registered in the triplet data storage unit 43. To do. Note that, as the number of triplets stored as basic data increases, the accuracy of the user determination data increases. In this embodiment, triplet data having at least 7000 or more triplets is used as basic data.

  The triplet number calculation unit 59 calculates the number of triplets made of the same character string. Specifically, the triplet number calculation unit 59 reads all triplets included in the basic data acquired by the basic data acquisition unit 58, and for each triplet of different character strings, the number of occurrences of triplets of the same character string as the triplet is calculated. It is counted and calculated as the triplet number.

  The weight value setting unit 60 sets a weight value for each triplet. In this embodiment, the weight value setting unit 60 acquires the triplet number calculated by the triplet number calculation unit 59 for each triplet, and sets the weight value according to the triplet number. Specifically, a triplet with a large number of triplets gives a higher weight value because it is more likely to indicate the characteristics of a registered user more accurately and stably, and a triplet with a small number of triplets is less stable as a parameter. Give a low weight value.

  In the present embodiment, a triplet having a triplet number of 11 or more is given a weight value of “1”, and a triplet having a triplet number of 10 or less is given a weight value of “0”. However, the present invention is not limited to this, and a weight value proportional to the number of triplets may be given.

  The triplet re-sorting unit 62 re-sorts triplet data suitable as user determination data. In the present embodiment, the triplet re-sorting unit 62 determines whether or not the weight value set by the weight value setting unit 60 is greater than or equal to a predetermined threshold value. Then, triplet data in which the weight value is equal to or greater than a predetermined value is selected as information suitable for user determination data. On the other hand, triplet data whose weight value is smaller than a predetermined value is excluded as information inappropriate for user determination data.

The variation coefficient calculation unit 61 calculates a variation coefficient (standard deviation / average value) of the first ratio value (R ₁ ) and the second ratio value (R ₂ ) for each triplet. Specifically, the coefficient of variation calculation unit 61 acquires the first ratio value (R ₁ ) and the second ratio value (R ₂ ) for each triplet re-sorted by the triplet re-sorting unit 62, and each ratio value Calculate the mean and standard deviation. A value obtained by dividing the standard deviation by the average value is calculated as a coefficient of variation (standard deviation / average value).

  Further, in the present embodiment, the triplet re-sorting unit 62 executes re-sorting by a coefficient of variation in addition to the above-described re-sorting by the weight value. Specifically, the triplet re-sorting unit 62 determines whether or not the variation coefficient calculated by the variation coefficient calculation unit 61 is equal to or less than a predetermined threshold. Then, triplet data having the variation coefficient equal to or smaller than a predetermined value is re-sorted as user determination data. On the other hand, triplet data in which the variation coefficient is larger than a predetermined value is determined to be inappropriate as user determination data, and is excluded. This is because the larger the variation coefficient, the larger the variation of each ratio value, and the lower the reliability as the data for user determination.

  The user determination data registration unit 63 registers user determination data in the user determination data storage unit 42. Specifically, the user determination data registration unit 63 registers only the triplet data rescreened by the triplet rescreening unit 62 from the basic data in the user determination data storage unit 42 as user determination data. It has become.

  Next, each component that realizes the user determination function will be described. As shown in FIG. 1, this user determination function mainly includes a determination target data acquisition unit 64, an evaluation value calculation unit 65, a feature value calculation unit 66, a user determination unit 67, and a post-processing execution unit 68 as a computer. It is realized by making the function.

  Hereinafter, each component that realizes the user determination function will be described in more detail.

  The determination target data acquisition unit 64 acquires determination target data for determining whether or not the user operating the key is a registered user. In the present embodiment, the determination target data acquisition unit 64 sequentially acquires triplet data for one determination from the triplet data storage unit 43 as determination target data among the triplet data sampled for the registered user to be determined. . In the present embodiment, the determination target data acquisition unit 64 considers a balance between real-time constant monitoring and processing speed, and determines the triplet data including 50 to 400 triplets as a determination target for one determination. It comes to get as data.

The evaluation value calculator 65 calculates an evaluation value (V) for each triplet. The evaluation value calculation unit 65 calculates the first ratio value based on the user determination data stored in the user determination data storage unit 42 for each triplet included in the determination target data acquired by the determination target data acquisition unit 64. An evaluation value (V) of (R ₁ ) and the second ratio value (R ₂ ) is calculated.

  Specifically, the evaluation value calculation unit 65 refers to all triplets (same triplet group) of the same character strings as the triplets included in the determination target data in the user determination data in the user determination data storage unit 42. . Then, the ratio value of the triplet included in the determination target data is compared with each ratio value of the same triplet group, and the rank relating to the ratio value of the triplet is obtained. A higher evaluation value (V) is calculated as the ranking is closer to the median value, and a lower evaluation value (V) is calculated as the ranking is closer to the highest value or the lowest value.

  FIG. 5 is a graph showing an evaluation method by the evaluation value calculation unit 65. Suppose that the ratio value is plotted on the horizontal axis and the triplet number is plotted on the vertical axis, the distribution shows the ratio value displayed as a solid line. In this case, when converted into a rank regarding the ratio value, a median value in the rank is obtained. Therefore, as shown by the dotted line in the figure, the triplet having the ratio value that is the median value is assigned the highest evaluation value, and the evaluation value is set to decrease as the rank goes away from the median value. Yes.

Moreover, in this embodiment, the evaluation value calculation part 65 calculates the evaluation value (V) according to the order | rank mentioned above by using following formula (1).
V (R) = a−b × | (LS) | × (L + S) Expression (1)
However,
a, b: Coefficient R: First ratio value or second ratio value L: Among triplets in the user determination data, the same character string as the triplet of the determination target data, and a ratio value larger than the ratio value of the triplet S: Number of triplets in the user determination data having the same character string as the determination target data triplet and having a ratio value less than or equal to the ratio value of the triplet

  In the present embodiment, the evaluation value (V) corresponding to the above-described rank is adopted in consideration of the distortion of the triplet ratio value distribution. However, the present invention is not limited to this. For example, as will be described later, when the ratio value distribution is close to a normal distribution, the standard deviation may be used as a reference, and the distance from the average value may be adopted as the evaluation value (V).

  The feature value calculation unit 66 calculates a feature value (C) related to the determination target data. This feature value (C) serves as an index for determining whether the determination target data is derived from the registered user. In the present embodiment, the feature value calculation unit 66 acquires each evaluation value (V) calculated by the evaluation value calculation unit 65 for all triplets included in the determination target data. Based on these evaluation values (V), the characteristic value (C) of the determination target data is calculated.

Specifically, the feature value calculation unit 66 calculates the feature value (C) by the following equation (2).
C = {Σ [W × V (R)]} / N (2)
However,
W: Weight value determined based on the number of triplets of the same character string as the triplet of the determination target data among the triplets in the user determination data N: Number of triplets of the unit to be evaluated

  In the present embodiment, the weight value (W) in the above equation (2) is set in the same manner as the weight value calculated by the weight value setting unit 60 described above. Specifically, for triplets with many identical triplets in the data for user determination, a high weight value is set because there is a high possibility that the characteristics of the determination target data are more accurately and stably indicated, and for triplets with few identical triplets. Is not stable as a parameter, so a low weight value is set.

Moreover, even if the number of triplets (N) of the unit to be evaluated in the above formula (2) is one triplet, the evaluation value for the first ratio value (R ₁ ) and the evaluation for the second ratio value (R ₂ ) If a value is used, N = 2.

  The user determination unit 67 determines whether the determination target data is derived from the registered user. In the present embodiment, the user determination unit 67 acquires the feature value (C) calculated by the feature value calculation unit 66, and compares the magnitude relationship between the feature value (C) and a predetermined cutoff value. When the feature value (C) is larger than the cutoff value, it is determined that the determination target data is derived from the registered user. On the other hand, when the feature value (C) is equal to or less than the cutoff value, it is determined that the determination target data is not input by the registered user.

  The cut-off value is set based on a balance between the probability of correctly determining a registered user as a registered user (specificity: specificity) and the probability of correctly determining another person different from the registered user as a stranger (sensitivity: sensitivity). Is done. Since this embodiment is based on the premise of continuous user authentication by constant monitoring, the cutoff value is set so that the specificity is high, that is, the probability that the registered user is erroneously determined as another person is low.

  The post-processing execution unit 68 executes predetermined post-processing according to the determination result of the user determination unit 67. In the present embodiment, when the user determination unit 67 determines that the post-processing execution unit 68 is a registered user, the post-processing execution unit 68 permits the key operation by the user as it is. On the other hand, when the user determination unit 67 determines that the user is not a registered user, predetermined post-processing is executed.

  As specific post-processing, a warning report to that effect may be issued to the administrator of the system 1, or nothing may be done until the input data is questionable. . In the above case, the key operation itself by the user is continued. On the other hand, the post-processing execution unit 68 may separately request the user who is operating the key to input a password or perform fingerprint authentication, or may prohibit the key operation. However, since it is difficult to set the determination accuracy to 100%, a sufficient ground is necessary to prohibit the key operation in consideration of the circumstances in the medical field.

  Next, regarding the operation of the user authentication system 1 executed by the user authentication program 1a of the present embodiment, each of the triplet data collection function, the user determination data creation function, and the user determination function will be described with reference to the drawings.

Operation by “triplet data collection function” FIG. 6 is a flowchart showing processing executed by the triplet data collection function. First, when triplet data as basic data or determination target data is collected by the user authentication system 1 of the present embodiment, the user waits for the user to operate the keyboard as the input means 2. When the keyboard is operated, the key operation information acquisition unit 51 acquires the key operation information (step S1). In the present embodiment, when triplet data for basic data is collected, the user ID is prompted to be input at the start of key operation, and consent is confirmed for collection of key operation information.

  Subsequently, the triplet extraction unit 52 extracts a triplet consisting only of lowercase alphabetic keys (az) from the key operation information (step S2). This eliminates triplets that are likely to be input with a rhythm that is different from the user's original input rhythm, such as when the shift key is pressed to input capital letters or the space key is pressed to convert kanji. This improves authentication accuracy.

Next, for each extracted triplet, the time data acquisition unit 53 acquires various time data (D ₁ , D ₂ , D ₃ , T ₁ , T ₂ , T _A ) shown in FIG. 4 (step S3). ), The triplet selection unit 54 selects an effective triplet based on the time data (step S4).

Specifically, as shown in FIG. 7, the pressing time (D ₁ , D ₂ , D ₃ ) of each key (first key, second key, third key) and each waiting time (T ₁ , T _2). ) and the respective total triplet time _{(T a)} is, determines whether or not included within the predetermined time limit (step S41 to S43). Then, only triplets in which all the time data are included within the time limit are selected (steps S41 to S43: all YES). On the other hand, if any time data deviates from the time limit (steps S41 to S43: any NO), the triplet is excluded (step S44).

  As a result, when the key operation is accidentally delayed or interrupted, any time data exceeds the upper limit of the time limit, and thus the triplet is excluded. Also, for triplets that are input at a speed higher than the user's original input speed because they are frequently input character strings such as their own names, any time data will have the lower limit of the time limit. Because it is below, it is excluded as well. Even if unnatural time data is acquired due to the OS (Operating System) of the computer, the triplet is excluded because it deviates from the time limit.

As described above, since only the triplets having a high probability of showing the typing habits and characteristics of the registered user are selected, the authentication accuracy is improved. In the present embodiment, in order to increase the selection accuracy, each key pressing time (D ₁ , D ₂ , D ₃ ), each waiting time (T ₁ , T ₂ ), and the total triplet time (T _A ). Although it is determined, the present invention is not limited to this, and only the pressing time (D ₁ , D ₂ , D ₃ ) or the waiting time (T ₁ , T ₂ ) of each key is determined and selected. You may do it.

Subsequently, the ratio value calculation unit 55 calculates the first ratio value (R ₁ = T ₁ / T ₂ ) and the second ratio value (R ₂ = D ₂ / D ₃ ) for each selected triplet (step) S5). As the second ratio value, the first key depression time _{(D 1)} and the ratio of the second key depression time _{(D 2) (D 1 /} D 2), or the first key depression time _(D 1) And the ratio (D ₁ / D ₃ ) of the pressing time (D ₃ ) of the third key may be calculated.

Subsequently, each triplet ratio value is calculated by extracting the sequence run Damaizu unit 56 sequentially and temporarily stored in the storage means 4 RAM. When a predetermined number of triplet groups are accumulated, the extraction order by the triplet extraction unit 52 is randomly rearranged (step S6). This prevents a heavy load on the computer or network for each keystroke. Also, by randomizing the extraction order, it becomes difficult to reproduce the character string of the key operation information, so that leakage of highly confidential information is prevented and security is improved.

Then, the triplet data registration unit 57 registers the first ratio value (R ₁ ) and the second ratio value (R ₂ ) in association with each triplet in the triplet data storage unit 43 as triplet data (step S7).

  The triplet data obtained as described above is used as basic data for creating user determination data or determination target data for determining a user who operates a key. Therefore, by adopting the ratio value as the triplet data, the influence of the variation in the type speed on the determination result is reduced. Further, in conjunction with the above-described randomization processing, a clue for reproducing the character string of the key operation information is eliminated, thereby enhancing the security aspect.

  In this embodiment, the randomization process (step S6) of the extraction order is executed after the ratio value calculation step (step S5). However, the present invention is not limited to this, and the triplet data registration step (step S7) is performed. If it is before, it may be executed before the ratio value calculating step (step S5).

Action by “User Determination Data Creation Function” FIG. 8 is a flowchart showing processing executed by the user determination data creation function. First, when creating user determination data by the user authentication system 1 of the present embodiment, the basic data acquisition unit 58 uses the triplet data collected for the registered user who wants to create user determination data as the basic data. Obtained from the data storage unit 43 (step S11).

  Next, the triplet number calculation unit 59 reads all triplets included in the basic data, and calculates the number of triplets for each triplet of the same character string (step S12). Then, the weight value setting unit 60 sets a weight value corresponding to the number of triplets (step S13).

  Subsequently, the triplet re-sorting unit 62 acquires the weight value assigned to each triplet and determines whether or not the weight value is equal to or greater than a predetermined threshold value (step S14). Then, the triplet data having the weight value equal to or greater than the predetermined value is rescreened as user determination data (step S14: YES). On the other hand, when the weight value is smaller than the predetermined value (step S14: NO), the triplet data is excluded as being unsuitable as user determination data (step S15). As a result, a triplet having a large number of triplets is employed because it is highly likely to show the characteristics of the registered user more accurately and stably. On the other hand, triplets with a small number of triplets are excluded because they are not stable as parameters.

Next, for each triplet data reselected by the weight value, the variation coefficient calculation unit 61 obtains an average value and a standard deviation of the first ratio value (R ₁ ) and the second ratio value (R ₂ ), and the variation A coefficient (standard deviation / average value) is calculated (step S16). Then, the triplet re-sorting unit 62 determines whether or not the variation coefficient is equal to or less than a predetermined threshold (step S17).

  As a result of this determination, if the variation coefficient is equal to or smaller than the predetermined value (step S17: YES), the user determination data registration unit 63 registers the triplet data as user determination data in the user determination data storage unit 42 ( Step S18). On the other hand, when the variation coefficient is larger than the predetermined value (step S17: NO), the triplet data is excluded as being unsuitable as user determination data (step S15). As a result, triplet data indicating the original characteristics of the registered user is employed as the user determination data with small variations in the ratio values.

  The sorting process from step S12 to step S18 is repeated for all triplet data (step S19). By executing the user determination data creation function as described above, user determination data suitable for use as a reference for determining whether or not the user is a registered user is created.

Action by “User Determination Function” FIG. 9 is a flowchart showing processing executed by the user determination function. First, when it is determined by the user authentication system 1 of this embodiment whether or not the user who is operating the key is the registered user himself / herself, the determination target data acquisition unit 64 collects one of the users to be determined. The triplet data for the number of times of determination is acquired from the triplet data storage unit 43 as determination target data (step S21).

Next, the evaluation value calculation unit 65 uses the first ratio value (R ₁ ) and the second value for each triplet included in the determination target data based on the user determination data stored in the user determination data storage unit 42. An evaluation value (V) for the ratio value (R ₂ ) is calculated (step S22). In the present embodiment, the evaluation value calculation unit 65 refers to all triplets of the same character string as the determination target triplet in the user determination data, and obtains a rank regarding the ratio value of the determination target triplet. A higher evaluation value (V) is assigned as the ranking is closer to the median value, and a lower evaluation value (V) is assigned as the ranking is closer to the highest value or the lowest value.

  Thereby, even when the ratio value distribution of each triplet is extremely distorted, an accurate evaluation value is given. This point will be described in more detail. First, as shown in FIG. 10, when the ratio value distribution is close to the normal distribution, according to the distance from the average value based on the standard deviation as in the conventional case. There is no problem even if the set evaluation function f (x) is used.

  In this case, a high evaluation value is given in the vicinity of the average value because it frequently appears as a ratio value of a triplet input by the user, and a low evaluation value is given because it is a ratio value that does not appear much as the distance from the average value increases. Is done. As a result, as shown in FIG. 11, since there is a large difference between the probability distribution P (x) of the registered user and the probability distribution Q (x) of others, the expected value 登録 of the evaluation value of the registered user for one triplet (F (x) · P (x)) is larger than the expected value 他 (f (x) · Q (x)) of the other person's evaluation value, and a correct evaluation can be made.

  On the other hand, as shown in FIG. 12, when the ratio value distribution is extremely distorted, a problem arises when the above-described evaluation function f (x) is used. Specifically, the data indicated by the black circles in the figure is intended to give a low evaluation because it is the end of the distribution. However, since the ratio value distribution of FIG. 12 is a distribution with a peak from the left and extending to the right, the standard deviation and the average value are the same as the ratio value distribution of FIG. For this reason, when the evaluation function f (x) is used, since the black circle data is close to the average value vicinity, a high evaluation value is given.

  On the other hand, according to the evaluation value calculation unit 65 of the present embodiment, as described with reference to FIG. 5, the highest evaluation value is assigned to the triplet having the ratio value that is the median of the ranks regarding the ratio value. In addition, a low evaluation value is given to a triplet whose rank is far from the median. Therefore, even for the black circle data, even if it is close to the average value, it is in the vicinity of the tail as the rank, and therefore a correctly low evaluation value is given.

  Returning to FIG. 9, when the evaluation value (V) of each triplet is calculated by the evaluation value calculation unit 65 (step S22), the feature value calculation unit 66 calculates the evaluation values (V of all triplets included in the determination target data). ) And the feature value (C) of the determination target data is calculated (step S23). Then, the user determination unit 67 compares the feature value (C) with a predetermined cutoff value (step S24).

  As a result of this comparison, when the user determination unit 67 determines that the feature value (C) is equal to or lower than the cutoff value (step S24: NO), the determination target data is key-operated by another person different from the registered user. It is determined that this has been done (step S25). Thereby, the post-processing execution unit 68 executes post-processing such as issuing a warning report to that effect to the administrator of the system 1 (step S26). Then, the process returns to step S21, and as long as the key is operated, the key operation information is constantly monitored and the user authentication process is continuously executed. If the key operation is prohibited, this flow is terminated.

  On the other hand, when the user determination unit 67 determines that the feature value (C) is greater than the cutoff value (step S24: YES), it is determined that the determination target data is derived from the registered user. (Step S27). Also in this case, the process returns to step S21, and the determination target data acquisition unit 64 acquires new determination target data. As long as the registered user operates the key, the key operation information is constantly monitored, and the user authentication process is continuously executed.

According to this embodiment as described above,
1. It is possible to constantly monitor key operation information from the keyboard and continuously execute user authentication processing.
2. Only data suitable for user determination data and determination target data can be selected from the key operation information, and the authentication accuracy can be improved.
3. It is possible to improve security by preventing key operation information from being analyzed.
4). Even when the ratio value distribution is extremely distorted, there is an effect that an accurate evaluation value can be given to each triplet.

  Next, specific examples of the user authentication program 1a and the user authentication system 1 according to the present invention will be described. In the first embodiment, assuming that the user authentication system 1 according to the present invention is applied to an information system in a hospital, eight subjects (28 to 61 years old) consisting of researchers and hospital employees, User determination data was created, and an experiment was performed to determine other person's determination target data based on the user determination data.

  First, in the first embodiment, the user authentication program 1a according to the present invention is installed in the personal computer PC (Personal Computer) of each subject, and the triplet data is collected in daily work using the above-described triplet data collection function. Collected. In Example 1, it took 4 to 40 days to collect sufficient basic data (10,000 triplets).

  The user authentication program 1a according to the first embodiment is written using the programming language C ++ (VisualStudio2005: registered trademark) and operates in the background of a Windows PC (Windows 98, Windows 2000 and Windows XP: registered trademarks). is there. In the first embodiment, the user authentication program 1a is configured to collect all key operation information by setting a global hook in the dynamic link library. The collected triplet data was accumulated in a local hard disk or NAS (Network Attached Storage) connected to a hospital network.

In the first embodiment, when the triplet data collection function is executed, the lower limit is set for the time for pressing each key (D ₁ , D ₂ , D ₃ ) and the time limit for each waiting time (T ₁ , T ₂ ). The value was set to 5 ms and the upper limit value was set to 250 ms. For the total triplet time (T _A ), only the upper limit value was set to 600 ms, and the triplet sorting unit 54 performed sorting. Moreover, the triplet processing number for one time by the extraction order randomizing unit was set to 100 to 300 triplets.

  In executing the user determination data creation function, 7000 triplets of basic data were read. The weight value setting unit 60 assigns “1” as a weight value to triplets having 11 or more triplets, and assigns “0” as a weight value to 10 or less triplets. The triplet re-sorting unit 62 re-sorts only triplets whose weight values are positive and whose ratio coefficient of variation is 0.357 or less. The limit value of the coefficient of variation was determined by trial and error by the present inventors.

  Furthermore, when executing the user determination function, the evaluation value calculation unit 65 calculates an evaluation value using the above equation (1). Note that a = 7 and b = 11 were used as the coefficients in the above equation (1). In addition, the feature value calculation unit 66 calculates the feature value using the above equation (2). As the weight value in the above formula (2), “1” is assigned to a triplet having 11 or more triplets, and “0” is assigned to 10 or less triplets.

  As described above, in the first embodiment, first, user determination data related to a subject is created from basic data (7000 triplet) of a subject. On the other hand, by subdividing basic data of a subject different from the subject as determination target data (test length: 150 triplets) and determining each determination target data, a so-called virtual “imposters '(imposters' challenge) ”. And this "other person challenge" was repeated about all the other subjects, and the result was integrated and it was set as the histogram of "other person" of FIG.

  Further, the virtual identity authentication is determined after the user determination data is created from the basic data of the subject. In this case, the number of triplets of the basic data is adjusted to be the same in both series so that the same part as the determination target data is not included in the basic data and the fair learning condition is satisfied. Then, these results are integrated into a histogram of “person” in FIG. In FIG. 13, normalization is performed with the peaks of each histogram.

  As shown in FIG. 13, in the personal authentication, it can be confirmed that there is a tendency for a higher feature value than other people to appear. However, since the histograms of “person” and “others” both extend to the vicinity of the right end, it is difficult to increase the sensitivity (probability of correctly determining others as others). On the other hand, on the left side of the graph, the end of each histogram is separated, indicating that the specificity (probability of correctly determining a registered user as a registered user) can be extremely high.

  As described above, according to the first embodiment, in the user authentication program 1a and the user authentication system 1 according to the present invention, it is possible to increase specificity by appropriately setting a cutoff value, that is, a registered user. It has been shown that the probability of being mistakenly determined as someone else can be reduced.

  Next, in the second embodiment, in order to analyze the performance of the user authentication program 1a and the user authentication system 1 according to the present invention, in the above-described first embodiment, the number of triplets of determination target data to be determined at one time is changed. ROC (Receiver Operating Characteristics) curves were created.

  Here, the ROC curve means that when “positive” is determined to be “other”, as shown in FIG. 14, the horizontal axis indicates “false positive rate (1-specificity): It is a graph plotted when “probability of determination” is taken and “sensitivity (probability of correctly determining another person as another person)” is taken on the vertical axis. This ROC curve shows the sensitivity and false positive rate at all selectable cut-off values, so that the user judgment performance at any cut-off value can be known.

  Further, the ROC curve changes depending on how the histograms of “person” and “others” as shown in FIG. 13 overlap. Specifically, if both histograms match exactly, the false positive rate and sensitivity are always equal. For this reason, as shown in FIG. 14, the ROC curve coincides with the diagonal line from the lower left (false positive rate: 0%, sensitivity 0%) to the upper right (false positive rate: 100%, sensitivity 100%). Shows no ability.

  On the other hand, the smaller the degree of overlapping of the histograms, the higher the sensitivity and the lower the false positive rate. Therefore, as shown in FIG. 14, the ROC curve has an upper left corner (false positive rate: 0%, sensitivity 100%). Drawn close to. Therefore, the closer the ROC curve is to the upper left, the higher the user determination performance.

  FIG. 15 shows ROC curves when the number of triplets of the determination target data is set to 50, 100, 150, and 200 in the second embodiment. As shown in FIG. 15, it is shown that the ROC curve shifts to the upper left as the number of triplets of the determination target data increases, and exhibits higher user determination performance.

  Here, the user judgment performance required for the user authentication system 1 according to the present invention is examined. First of all, the conventional login authentication at the start of use requires a very low “false negative rate (probability of judging another person as the wrong person)”, while the relatively high “false positive rate (who is wrong as the wrong person). "Probability of determination)" is allowed. This is because even if the principal is rejected by mistake, it can be retried immediately. On the other hand, once another person is mistakenly authenticated, there is no opportunity for subsequent authentication. Therefore, conventionally, when evaluating a system for authenticating a user, attention is focused on a portion with high sensitivity and low specificity in the ROC curve.

  However, as in the present invention, when aiming at continuous user authentication processing by continuous monitoring, it is important that the false positive rate (the probability of erroneously determining the person as another person) is low. This is because, for example, even in an extremely excellent authentication system with a false positive rate of 0.01%, in an organization of 500 people, if everybody receives an average of 100 user authentications per day, an average of 5 times a day This is because a false alarm is issued and the usability is poor. On the other hand, in the present invention, unlike the conventional login authentication, there is no problem even if the false negative rate (the probability of erroneously determining another person as the person) is somewhat high. This is because the user authentication process is repeatedly executed as long as the key operation is performed, and it is only necessary to be able to determine the other person by any one of the determination processes.

  As described above, in the always-monitoring type user authentication system 1, a portion with a low false positive rate (probability that a person is erroneously determined to be another person) is particularly important. For this reason, FIG. 16 shows an enlarged scale of only the horizontal axis in FIG. As shown in FIG. 16, in the second embodiment, for example, when the number of triplets of the determination target data is 150, an extremely low false positive rate of 0.5% is achieved at an allowable false negative rate of 39%. Yes.

  As described above, according to the second embodiment, the user authentication program 1a and the user authentication system 1 according to the present invention exhibit higher user determination performance as the number of triplets of determination target data increases, and allowance. It has been shown that a high specificity can be obtained at a possible false negative rate. Therefore, the user authentication program 1a and the user authentication system 1 according to the present invention are user authentication technologies that are applied to information systems such as hospitals that handle document data that is not allowed to be “spoofed” or tampered by others, such as electronic medical records. It has been shown to be preferred.

  The user authentication program 1a and the user authentication system 1 according to the present invention are not limited to the above-described embodiment, and can be changed as appropriate.

  For example, in the above-described embodiment, all functions of the triplet data collection function, the user determination data creation function, and the user determination function are executed by the user authentication program 1a. However, the present invention is not limited to this. For example, each function may be an independent program as a triplet data collection program, a user determination data creation program, and a user determination program.

  In the above-described embodiment, the triplet data obtained from the key operation information is sequentially determined in real time. However, the present invention is not limited to this. For example, triplet data may be stored in association with the input key operation information, and the triplet data may be determined when there is doubt about the reliability of the key operation information.

It is a block diagram which shows embodiment of the user authentication program and user authentication system which concern on this invention. It is a figure which shows an example of the data for user determination of this embodiment. In this embodiment, it is a figure explaining the triplet extracted from key operation information. In this embodiment, it is a figure explaining the time data acquired about a triplet. In this embodiment, it is a graph which shows the evaluation method by the evaluation value calculation part 65. It is a flowchart figure which shows the process performed by the triplet data collection function in this embodiment. In this embodiment, it is a flowchart figure which shows the selection process by a triplet selection part. It is a flowchart figure which shows the process performed by the data creation function for user determination in this embodiment. It is a flowchart figure which shows the process performed by the user determination function in this embodiment. In this embodiment, it is a graph which shows the ratio value distribution close | similar to normal distribution, and the evaluation function on the basis of a standard deviation. FIG. 10 is a graph comparing the probability distribution P (x) of registered users with the probability distribution Q (x) of others. In this embodiment, it is a graph which shows the case where ratio value distribution is distorted extremely. In the present Example 1, it is a histogram which shows the characteristic value of "person" and "others". It is a graph for demonstrating a ROC curve. In Example 2, it is a graph which shows a ROC curve when the number of triplets of determination object data is changed. In FIG. 15, it is the graph which expanded only the scale of the horizontal axis 20 times.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 User authentication system 1a User authentication program 2 Input means 3 Display means 4 Storage means 5 Arithmetic processing means 41 Program storage part 42 User determination data storage part 43 Triplet data storage part 51 Key operation information acquisition part 52 Triplet extraction part 53 Time data Acquisition unit 54 Triplet selection unit 55 Ratio value calculation unit 56 Extraction order randomization unit 57 Triplet data registration unit 58 Basic data acquisition unit 59 Triplet number calculation unit 60 Weight value setting unit 61 Fluctuation coefficient calculation unit 62 Triplet reselection unit 63 User judgment Data registration unit 64 Determination target data acquisition unit 65 Evaluation value calculation unit 66 Feature value calculation unit 67 User determination unit 68 Post-processing execution unit

Claims (18)

A triplet consisting of a combination of any three consecutive press keys obtained from the operation information of the keyboard operated by the registered user, and from the time when the first key is pressed to the time when the second key is pressed among the operation information of this triplet. first rate value determined as a ratio of the second waiting time from the first wait time (T ₁₎ a second key is pressed until the third key is pressed (T ₂₎ and (R ₁₎ at least A user determination data storage unit that stores in advance user determination data associated with the registered user;
A key operation information acquisition unit for continuously acquiring key operation information when the keyboard is operated;
A triplet extraction unit for extracting a predetermined triplet from the key operation information;
A time data acquisition unit for acquiring the first waiting time (T ₁ ) and the second waiting time (T ₂ ) for each triplet extracted by the triplet extraction unit;
A ratio value calculation unit for calculating the first ratio value (R ₁ ) for each triplet;
A triplet data registration unit for registering in the triplet data storage unit as triplet data in which the first ratio value (R ₁ ) is associated with each triplet;
A determination target data acquisition unit that acquires triplet data for one determination out of the triplet data registered in the triplet data storage unit as determination target data;
An evaluation value calculation unit that calculates an evaluation value (V) of each first ratio value (R ₁ ) in the determination target data based on the user determination data stored in the user determination data storage unit;
A feature value calculation unit that calculates a feature value (C) related to the determination target data based on each evaluation value (V) calculated by the evaluation value calculation unit for all triplets included in the determination target data;
A computer is caused to function as a user determination unit that determines whether the determination target data is input by the registered user based on a magnitude relationship between the feature value (C) and a predetermined cut-off value. User authentication program. 2. The user determination data storage unit according to claim 1, wherein the second ratio value (R ₂ ) obtained as a ratio of pressing times of any two pressing keys among three consecutive pressing keys constituting the triplet is a user. It is further stored as judgment data,
The time data acquisition unit acquires the pressing time (D ₁ , D ₂ , D ₃ ) of each key for each triplet extracted by the triplet extraction unit,
The ratio value calculation unit calculates the second ratio value (R ₂ ) for each triplet,
The triplet data registration unit associates the second ratio value (R ₂ ) with each triplet and registers it as triplet data in the triplet data storage unit,
The evaluation value calculation unit calculates an evaluation value (V) for the second ratio value (R ₂ ) in the determination target data based on the user determination data stored in the user determination data storage unit. ,
The feature value calculation unit calculates a feature value (C) related to the determination target data based on an evaluation value (V) for the first ratio value (R ₁ ) and the second ratio value (R ₂ ). A user authentication program for causing a computer to function. 3. The first waiting time (T ₁ ) and the second waiting time (T ₂ ) to the key pressing time (D ₁ , D ₂ ) in each triplet acquired by the time data acquisition unit. D ₂ , D ₃ ) have a triplet sorting unit that sorts triplets within a predetermined time limit,
For each triplet selected by the triplet selection unit, the ratio value calculation unit calculates a second ratio value (R ₂ ) that is a ratio between the first ratio value (R ₁ ) and the pressing time of each key. ) And the triplet data registration unit causes the computer to function such that both are associated with each other and registered as triplet data in the triplet data storage unit.   The evaluation value calculation unit according to any one of claims 1 to 3, wherein a ratio value of a triplet included in the determination target data and a ratio value of the same triplet group stored in the user determination data To obtain a rank regarding the ratio value of the triplet, the higher the rank is closer to the median value, the higher the evaluation value (V) is calculated, and the lower the rank is, the lower the evaluation value (V) is. A user authentication program for causing a computer to function to calculate. 5. The user authentication program according to claim 4, wherein the evaluation value calculation unit causes a computer to function so as to calculate the evaluation value (V) by the following equation (1).
V (R) = a−b × | (LS) | × (L + S) Expression (1)
However,
a, b: Coefficient R: First ratio value or second ratio value ( ratio of pressing times of any two pressing keys among three consecutive pressing keys constituting a triplet)
L: Number of triplets in the user-determination data having the same character string as that of the determination target data and having a ratio value larger than the ratio value of the triplet S: Of triplets in the user-determination data The number of triplets having the same character string as the triplet of the determination target data and having a ratio value equal to or lower than the ratio value of the triplet 6. The user authentication program according to claim 5 , wherein the feature value calculation unit causes a computer to function so as to calculate the feature value (C) by the following equation (2).
C = {Σ [W × V (R)]} / N (2)
However,
W: Weight value determined based on the number of triplets of the same character string as the triplet of the determination target data among the triplets in the user determination data N: Number of triplets of the unit to be evaluated In any one of Claims 3-6,
The time data acquisition unit further acquires, for each triplet extracted by the triplet extraction unit, a total triplet time from when the first key is pressed until the third key is released,
The user authentication program, wherein the triplet selection unit causes a computer to select only triplets within a predetermined time limit from the acquired total triplet time. In any one of Claims 1-7,
Before registering the triplet data of the triplet extracted by the triplet extraction unit in the triplet data storage unit, the triplet data is temporarily stored in the RAM sequentially in accordance with the extraction order, and when a predetermined number of triplet groups are stored, A user authentication program that causes a computer to function as an extraction order randomizing unit that randomly rearranges the order of each triplet stored in a RAM . In any one of Claims 1-8,
A basic data acquisition unit that acquires basic data for creating user determination data related to a predetermined registered user from the triplet data stored in the triplet data storage unit;
A triplet number calculating unit for calculating the number of triplets of the same character string included in the basic data;
A weight value setting unit that sets a weight value for each triplet according to the number of triplets calculated by the triplet number calculating unit;
A coefficient of variation calculation unit that calculates at least a coefficient of variation (standard deviation / average value) of the first ratio value (R ₁ ) for each triplet;
A triplet re-sorting unit that re-sorts only triplet data in which the weight value set by the weight value setting unit is greater than or equal to a predetermined value and the variation coefficient calculated by the variation coefficient calculation unit is less than or equal to a predetermined value as the user determination data A user authentication program characterized by causing a computer to function. A triplet consisting of a combination of any three consecutive press keys obtained from the operation information of the keyboard operated by the registered user, and from the time when the first key is pressed to the time when the second key is pressed among the operation information of this triplet. first rate value determined as a ratio of the second waiting time from the first wait time (T ₁₎ a second key is pressed until the third key is pressed (T ₂₎ and (R ₁₎ at least A user determination data storage unit that stores in advance user determination data associated with the registered user;
A key operation information acquisition unit for continuously acquiring key operation information when the keyboard is operated;
A triplet extraction unit for extracting a predetermined triplet from the key operation information;
A time data acquisition unit for acquiring the first waiting time (T ₁ ) and the second waiting time (T ₂ ) for each triplet extracted by the triplet extraction unit;
A ratio value calculation unit for calculating the first ratio value (R ₁ ) for each triplet;
A triplet data registration unit for registering in the triplet data storage unit as triplet data in which the first ratio value (R ₁ ) is associated with each triplet;
A determination target data acquisition unit that acquires triplet data for one determination out of the triplet data registered in the triplet data storage unit as determination target data;
An evaluation value calculation unit that calculates an evaluation value (V) of each first ratio value (R ₁ ) in the determination target data based on the user determination data stored in the user determination data storage unit;
A feature value calculation unit that calculates a feature value (C) related to the determination target data based on each evaluation value (V) calculated by the evaluation value calculation unit for all triplets included in the determination target data;
A user determination unit that determines whether or not the determination target data is input by the registered user based on a magnitude relationship between the feature value (C) and a predetermined cut-off value. User authentication system. 11. The second ratio value (R ₂ ) obtained as a ratio of pressing times of any two pressing keys among the three consecutive pressing keys constituting the triplet is stored in the user determination data storage unit according to claim 10. It is further stored as judgment data,
The time data acquisition unit acquires the pressing time (D ₁ , D ₂ , D ₃ ) of each key for each triplet extracted by the triplet extraction unit,
The ratio value calculation unit calculates the second ratio value (R ₂ ) for each triplet,
The triplet data registration unit associates the second ratio value (R ₂ ) with each triplet and registers it as triplet data in the triplet data storage unit,
The evaluation value calculation unit calculates an evaluation value (V) for the second ratio value (R ₂ ) in the determination target data based on the user determination data stored in the user determination data storage unit. ,
The feature value calculation unit calculates a feature value (C) related to the determination target data based on an evaluation value (V) for the first ratio value (R ₁ ) and the second ratio value (R ₂ ). A featured user authentication system. In Claim 10 or Claim 11, the first waiting time (T ₁ ) and the second waiting time (T ₂ ) in each triplet acquired by the time data acquisition unit or the key pressing time (D ₁ , D ₂ , D ₃ ) have a triplet sorting unit that sorts triplets within a predetermined time limit,
For each triplet selected by the triplet selection unit, the ratio value calculation unit calculates a second ratio value (R ₂ ) that is a ratio between the first ratio value (R ₁ ) and the pressing time of each key. ) A user authentication system characterized in that a second ratio value (R ₂ ) is calculated, and the triplet data registration unit associates the two and registers them as triplet data in the triplet data storage unit.   In any one of Claims 10-12, the said evaluation value calculation part is the ratio value of the triplet contained in the said determination object data, and the ratio value of the same triplet group memorize | stored in the said data for user determinations. To obtain a rank regarding the ratio value of the triplet, the higher the rank is closer to the median value, the higher the evaluation value (V) is calculated, and the lower the rank is, the lower the evaluation value (V) is. A user authentication system characterized by calculating. 14. The user authentication system according to claim 13, wherein the evaluation value calculation unit calculates the evaluation value (V) by the following equation (1).
V (R) = a−b × | (LS) | × (L + S) Expression (1)
However,
a, b: Coefficient R: First ratio value or second ratio value ( ratio of pressing times of any two pressing keys among three consecutive pressing keys constituting a triplet)
L: Number of triplets in the user-determination data having the same character string as that of the determination target data and having a ratio value larger than the ratio value of the triplet S: Of triplets in the user-determination data The number of triplets having the same character string as the triplet of the determination target data and having a ratio value equal to or lower than the ratio value of the triplet 15. The user authentication system according to claim 14 , wherein the feature value calculation unit calculates the feature value (C) by the following equation (2).
C = {Σ [W × V (R)]} / N (2)
However,
W: Weight value determined based on the number of triplets of the same character string as the triplet of the determination target data among the triplets in the user determination data N: Number of triplets of the unit to be evaluated In any one of Claims 12-15,
The time data acquisition unit further acquires, for each triplet extracted by the triplet extraction unit, a total triplet time from when the first key is pressed until the third key is released,
The user authentication system, wherein the triplet sorting unit sorts only triplets within a predetermined time limit from the acquired total triplet time. In any one of Claims 10-16,
Before registering the triplet data of the triplet extracted by the triplet extraction unit in the triplet data storage unit, the triplet data is temporarily stored in the RAM sequentially in accordance with the extraction order, and when a predetermined number of triplet groups are stored, A user authentication system comprising an extraction order randomizing unit that randomly rearranges the order of each triplet stored in a RAM . In any one of Claims 10-17,
A basic data acquisition unit that acquires basic data for creating user determination data related to a predetermined registered user from the triplet data stored in the triplet data storage unit;
A triplet number calculating unit for calculating the number of triplets of the same character string included in the basic data;
A weight value setting unit that sets a weight value for each triplet according to the number of triplets calculated by the triplet number calculating unit;
A coefficient of variation calculation unit that calculates at least a coefficient of variation (standard deviation / average value) of the first ratio value (R ₁ ) for each triplet;
A triplet re-sorting unit that re-sorts only triplet data in which the weight value set by the weight value setting unit is greater than or equal to a predetermined value and the variation coefficient calculated by the variation coefficient calculation unit is less than or equal to a predetermined value as the user determination data And a user authentication system.

Images