JPH0357082A - Personal certification device - Google Patents

Abstract

PURPOSE:To prevent decoding or duplicating by collating the output from a key touch feature extraction part with the data of a feature storage part and deciding whether identity is certificated or not. CONSTITUTION:A key touch input part 1, key touch feature extraction part 2, feature storage part 3 and feature collation part 4 are provided. A person to obtain certification inputs a keyword, which can be arbitrarily selected by the person himself in advance, by a keyboard and registers the feature of a method for key touch input, namely, an interval between touches for the adjacent characters of the keyword to the feature storage part 3. At the time of certification, the same keyword is inputted whether the identity is certificated or not is decided by investigating an error to the feature registered in advance. Thus, whether the identity is certificated or not can be exactly decided.

Description

[Detailed description of the invention] [Industrial application field]

TECHNICAL FIELD The present invention relates to a personal authentication device that distinguishes between the user himself and a stranger.

[Conventional technology]

Recently, in daily life and in the workplace, it has become common to have machines distinguish between the person in question and another person (so-called personal authentication). For example, authentication of a person attempting to operate an automatic teller machine (cash dispenser), authentication of a person attempting to operate a computer used jointly by multiple people (multi-user computer), This includes authentication of people leaving the site. Conventionally, devices that use a password, a personal identification number, an MS pattern, a fingerprint, or the like as an authentication means are known as devices for performing such personal authentication. Additionally, a device has been proposed in which a signature is signed using a pen-shaped coordinate input device, and the coordinates of the pen tip and pressure are used as authentication means (Nikkei Electronics magazine, October 3, 1988 (No. 457).
), P98).

[Ti, F[ which the invention seeks to solve] (problem) However, the authentication means used in the above-mentioned conventional personal authentication devices have problems in that they can be deciphered or copied, or are not suitable for the lifestyle habits of Japanese people. (Explanation of the problem) To avoid forgetting a password or PIN, it is best to associate it with something meaningful about the person. Therefore, there is a high possibility that it can be deciphered by making inferences based on the person's history, hobbies, etc. Retinal patterns and fingerprints can be taken from a person, reproduced, and read by a machine. Furthermore, Japan does not have a lifestyle in which individuals practice their signatures with good reproducibility as a means of authenticating themselves, and personal authentication devices that use signatures as a means of authentication are not suitable for Japanese people. The present invention aims to solve the above-mentioned problems. [Means to solve the problem]

In order to solve the above problems, the personal authentication device of the present invention includes:
In order to reduce the possibility of deciphering or copying, and to eliminate the need for special practice to ensure reproducibility, as with signatures, we decided to take the following measures. That is, the personal authentication device of the present invention includes a key touch input section for inputting a keyword manually, a key touch feature extraction section for extracting the tanochi spacing between adjacent characters of the keyword, and pre-storing data for each individual regarding the touch spacing. The device is configured to have a feature storage section for storing information about the identity of the person, and a feature matching section for comparing the output from the key feature extraction section with the data in the feature storage section to determine whether or not the person is the real person.

[Work I Individuals have their own unique way of moving their fingers when key-touching a specific word (keyword), and it is extremely difficult for others to imitate this. The present invention attempts to authenticate individuals by utilizing this point. First, an individual who wishes to obtain authentication enters a keyword that can be arbitrarily selected by the individual using a keyboard, and then memorizes the characteristics of the key touch input method, that is, the touch interval between adjacent characters of the keyword. Register with the department. At the time of authentication, the user is asked to enter the same keyword and check for discrepancies with the pre-registered features to determine if the person is the person. Determine whether or not. Since it is extremely difficult to imitate the sequence of finger movements that occur when inputting a specific word using a key touch, the identification of the person is extremely accurate. 【Example】

Hereinafter, embodiments of the present invention will be described in detail based on the drawings. FIG. 1 shows blocks of a personal authentication device according to an embodiment of the present invention. #Show precepts. In FIG. 1, 1 is a key touch input section, 2 is a key touch feature extraction section, 3 is a feature storage section,
4 is the feature matching section. The key touch human power section l is a keyboard that can input words and phrases.
Keyboards, which are widely used as input devices for word processors, etc., can be used. From here, a word or phrase arbitrarily determined by the individual or a word or phrase specified in advance by the certifier, that is, a keyword, is input. FIG. 2 is a diagram showing an example of key touch input, and this example shows the case where the word rPATENTJ is input as a keyword. The horizontal axis represents the input time based on the time when the first character "P" was input. Each character forming a keyword is key data for personal authentication, so
Let's call it key data. Figure 3 shows the relationship between the key data and the touch time at which it was input in the case of Figure 2. The key touch feature extraction unit 2 extracts key touch input features, that is, when inputting the keyword rPATENT, the key touch feature extraction unit 2 determines the time interval between adjacent characters when each character (key data) is input. Extract things. Figure 4 shows that if the input method is as in the example in Figure 2,
It is a diagram showing the relationship between adjacent key data and the time interval of their input touches. This diagram shows that it takes 0.2 seconds from inputting "P" to inputting "A", 0.3 seconds from inputting r A J to manually inputting "T", and After inputting rE
It takes 0.2 seconds to input "E" and 0.1 seconds to input r N J, and r N
This shows that the keyword rPATENTJ was inputted manually, as it took 0.3 seconds to input TJ after inputting J. It's almost the same when done manually, but it's different depending on the person.In other words, each person has their own unique way of moving their fingers when inputting a certain keyword, and personal authentication can be done by determining this. The feature storage unit 3 stores the features extracted by the key touch feature extraction unit 2 when first registering personal data regarding the keystroke input touch method in this personal authentication device. By the way, the feature matching section 4 matches the features stored in the feature storage section 3 with the features extracted by the key touch feature extraction section 2 when a key touch is performed manually for personal authentication. .Compared with the features stored in the feature storage unit 3, if the error between the two is smaller than a predetermined level, it is determined that the person is the person, and if it is greater than the predetermined level, it is determined that the person is a stranger.Next, personal authentication Let me explain a specific example. When Mr. A registers his own data in the personal authentication device, he selects the phrase rPATENTJ as a keyword,
Let's assume that you have registered by key-clicking input in the manner shown in Figure 2. In other words, the characteristics for determining Mr. A are:
Assume that a special image as shown in FIG. 4 is stored in the feature storage unit 3. Now, suppose that two people, Mr. A and Mr. B, each input the keyword rPATENTJ from the key touch input section 1 in order to obtain a personal L2 ID. The key touch feature extraction unit 2 extracts adjacent key data and touch interval data for two people as shown in FIG. Figure 5 (a) is Mr. A's data. Figure 5 (mouth) is B.
It shows the data of Mr. The feature matching unit 4 matches the degree of similarity between these data and the data in FIG. 4 by, for example, checking the root mean square error. Figure 6 shows the calculation and judgment results at that time. In the case of Mr. A, the root mean square error between the data in Figure 5 (a) and the data in Figure 4 is calculated to be 0.4. In the case of Mr. B, the root mean square error between the data in FIG. 5 (mouth) and the data in FIG. 4 is calculated to be 2.4. When inputting the same keyword rPATENTJ, Mr. B's sequence of finger movements is different from that of Mr. A, so the root mean square error of Mr. B will inevitably be large. Therefore, it is determined whether it is Mr. A or not. If the level is predetermined as the root mean square error -0.5, etc., when Mr. A inputs, the root mean square error will be smaller than the judgment level (0.5, etc.).
4≦0.5) When it is determined that Mr. A himself is j, and Mr. B inputs, the root mean square error is larger than the judgment level (
2.4 >0.5) r someone else (not Mr. A)”
It is determined that In addition, if the above judgment level is set low,
The exclusion rate for strangers will increase, but if it is too low, even the person may not be recognized as the person in question, so the judgment level should be determined appropriately so as not to be too low. In the above example, the degree of similarity is compared using the root mean square error, but the method of comparison is not limited to this, and methods such as taking the average of absolute values can be used to calculate the degree of similarity between two data strings. Any method is fine as long as it is suitable. An individual's key touch input habits may change little by little over the years, depending on their level of proficiency or familiarity with key touch. Therefore, if the standard data stored in the feature storage unit 3 remains unchanged, there is a possibility that the identity of the person cannot be authenticated even though the person is the person. Such situations can be dealt with by providing the personal authentication device with a learning function. That is, data regarding the new key performance when it is determined that the person is the real person is stored, and the contents of the feature storage unit 3 are updated (learned) little by little. For example, take the average of the data from the past three times when it was determined that the person is the real person, and use that as new reference data in the feature storage unit 3. By providing a learning function in this way, it is possible to maintain a high matching rate even if the person's habits change. A learning function can also be provided for the purpose of increasing the exclusion rate of others. For example, it is conceivable that someone else tries many times to obtain authentication for a certain individual and fails.
In such a case, the characteristics of the person who failed (relationship between adjacent key data and touch interval) should be memorized (learned), and when a human effort with those characteristics is performed, it will be judged as a "other person". It is to make it. This can increase the rate of exclusion of others. Note that if the number of characters in the keyword is small (for example, 4)
, there is a high possibility that the key touch input characteristics will be similar, but the possibility of similarity decreases as the number of characters increases. Therefore, in order to increase the matching rate, it is possible to increase the number of characters in the keyword, but if there are too many characters, the input operation during authentication becomes troublesome. Therefore, the number of characters in the keyword is appropriately determined as necessary, taking into consideration the number of people to be authenticated.

【Effect of the invention】

As described above, since the personal authentication device of the present invention uses manual key touch as the authentication means, it has the following effects. ■ There are individual differences in the way keytanchi works,
It is generally difficult for others to imitate it. Therefore, there is little chance of it being deciphered or copied. ■ A general-purpose keyboard such as a personal computer or word processor can be used as a key-touch input method, and no special device is required as a human-powered device for authentication. Therefore, it is possible to set up a personal authentication device at a low cost by using the equipment already in the office. ■ OA in the workplace, etc. (Ott1ce Auto@at
ion) is progressing rapidly, and there are more opportunities to become familiar with the keyboard, but since it is possible to select words and phrases that you have used manually at work etc. as keywords for authentication, it is necessary to practice specially for keywords. do not have.

[Brief explanation of drawings]

Figure 1: Block Ii of a personal authentication device according to an embodiment of the present invention Figure 2: An example of key touch human power Figure 3:
・Diagram 4 showing the relationship between key data and the time at which it was input...Diagram 5 showing the relationship between adjacent key data and the time intervals of their manual touches...Adjacent key data and touch intervals 2 about
FIG. 6 is a diagram showing human data...A diagram explaining an example of calculation and determination results in the feature matching section, l is a key touch input section, 2 is a key touch feature extraction section, and 3 is a feature writing section. The yearning section and 4 are the feature matching section.

Claims (1)

[Claims] a key touch input section for inputting a keyword; a key touch feature extraction section for extracting a touch interval between adjacent characters of the keyword; a feature storage section for storing in advance data for each individual regarding the touch interval; A personal authentication device comprising: a feature matching section that matches the output from the touch feature extracting section with data in the feature storage section to determine whether or not the person is the person in question.