DE102008041861A1 - Biometric identification method using feature vectors and analysis against multiple biometric samples - Google Patents

Abstract

The present invention relates to a method for identifying a user of a system by means of a biometric sample having user-typical features, wherein for each user of a plurality of users of at least one biometric sample of the user at least one feature vector (N) is stored in at least one vector space and in that a corresponding access feature vector is generated from an access probe for the user to be identified, wherein a comparison of the access feature vector with at least one feature vector of at least one vector space is performed for identification and / or the identification is an analysis of at least one access sample of one user against all or one part for all Users of the system deposited biometric samples.

Description

BACKGROUND OF THE INVENTION

FIELD OF THE INVENTION

The The present invention relates to a method of identification a user of a system using a biometric sample as well a corresponding device and a software product for implementation a corresponding procedure.

STATE OF THE ART

Various different identification systems for identifying persons are known from the prior art, for example in access control to buildings or to a factory premises or when logging into computers or computer networks. In addition to the identification via a simple password, biometric methods are known in which biometric data of a user are used for its identification. Biometric data can be, for example, fingerprints, the image of an eye, the distances of the eyes, ears, cheekbones etc. in the face, the voice of a human being or else the typing behavior when using keyboards, touch pads, key blocks, etc. Corresponding methods for identifying persons by means of the typing behavior are described, for example, in US Pat WO 98/06020 A2 or the US 6,151,593 A described.

Even though Such methods already provide good results, there is a further need, the method at the same time To simplify and make identification security more effective or at the same level of efficiency, a better identification security to reach. In particular, there is a need for the methods to improve so that a false identification is excluded becomes, with which an unauthorized user under wrong identity Access or access to a protected area would.

DISCLOSURE OF THE INVENTION

OBJECT OF THE INVENTION

It is therefore an object of the present invention, a method and to provide a device in which the above-mentioned Goals can be achieved and in particular an efficient one and secure identification method is given, which in particular also offers the possibility of misidentifications largely excluded.

TECHNICAL SOLUTION

These Task is solved by a method with the features of claim 1 or 19, a software product having the features of claim 25 and a device having the features of the claim 26. Advantageous embodiments are the subject of the dependent Claims.

The The above objects are according to a first aspect of the invention solved by the fact that for each user of a plurality by users of a system, for example, for each employee a company with access control to a building or a computer system, of at least one biometric sample the user stores at least one feature vector in at least one vector space with access control, a corresponding access feature vector is generated from an access sample and for identification Comparison of the access feature vector with at least one feature vector at least one vector space takes place.

With Such a procedure, for example, a kind of biometric Map or digital map for all users of one Systems can be created and a corresponding access sample can in a quick and easy way to this map d. H. the vector space be classified. Accordingly, in the method first at least one biometric sample for each user of the system detected for the at least one feature vector in at least a vector space is generated. The feature vector can hereby one-dimensional or multi-dimensional vector, one or more Features of a biometric sample contains. For example can in a Tipperkennung the feature vector and consequently also the access feature vector corresponding data over the Holding time of keys, the transition period, ie the transition from pressing one key to pressing the next Key or releasing one key to press the next one Key, the attack frequency, d. H. the pressing of buttons per unit time and the like included. The features can this include both directly collected data, z. B. so-called Raw tip data at tip detection, or edited and revised Data, allowing a variety of characteristics from a biometric Sample, such as a tip sample, can be determined.

When using the tipper identifier as a biometric method, the features may include, but are not limited to, the error rate or frequency, such as indicated by the use of the clear key, or the tip rate, which indicates how many words are written in a given time Tip rate can be determined by the number of typed blanks. In addition, information regarding the frequency of use of certain function keys, the selection of certain keys that are provided on the keyboard several times, for example, the use of the left or right shift key on a computer keyboard or other abnormalities such as time or key-related values, which fall within the usual framework as features are used. Other examples include the use of overrides that involve pressing the next button before releasing the previous button.

Next however, other biometric samples, as mentioned above, for example fingerprints, Eye and face pictures, characteristics of the voice etc. Find use.

The corresponding features of the feature vectors can be found in one or more one- or multi-dimensional vector spaces be provided.

For each user can have multiple feature vectors of multiple biometric Samples are stored.

Of the Vector space with the feature vectors can be segmented, where only feature vectors with certain coordinates are considered become. Similarly, in the vector space only Feature vectors of certain users are considered or the feature vectors are grouped into user groups, such as users at a particular location and the like.

Furthermore can be defined in the vector space areas, for example by the position of the feature vectors for a given one Users are given. A corresponding area can be exactly the feature vectors of the user or, moreover, a similarity area cover, d. H. include an area where similar Feature vectors can lie.

Corresponding can be for each feature vector and / or for several Feature vectors are defined a typical area in which similar feature vectors may be present.

at The identification based on the feature vectors can be the identification without knowledge of the alleged identity of the user. This means that not because of the given identity the access sample with a special stored reference sample is compared, but preferably against several Feature vectors, in particular different users.

Corresponding is for a further aspect of the invention for the self-employed and independent of the other aspects of the invention protection is sought, given a procedure in which the identification by analyzing a biometric access sample not only against a reference pattern of the alleged user takes place, but opposite all or part of the users or the deposited biometric Samples are done. This ensures that there are coincidental similarities of users and / or insufficient accuracy of the identification process lead to misidentifications. Rather, at a be determined such that possibly several User for the submitted biometric access sample in Question, in order to get further identification measures to be able to take. Accordingly, the analysis may in particular against biometric samples of at least two different ones Users.

Around but at the same time to maintain the efficiency of the process or To increase, can be a pre-selection of eligible users by comparing the access feature vector with deposited Feature vectors according to the first aspect of the present invention Invention be made.

By namely, the vector comparison is possible, for example a similarity range of the access feature vector define and check which feature vectors or which areas of deposited feature vectors in the similarity area of the access feature vector. Accordingly, then in a Downstream identification step a detailed analysis for the selected feature vectors or users respectively.

at the comparison of the feature vectors with the access feature vectors may also be a further limitation to that be made that only a comparison of the access feature vector with feature vectors of a user group, for example for those looking for a specific location or Be assigned to plant employees. By doing so leaves the efficiency also increases.

The comparison of the access feature vector with the feature vector may include the determination of the nearest feature vector, such that based on this feature vector, the access feature vector may be assigned to one or more user areas and / or one or more typical areas or neighborhoods to the typical area in the vector space. This can be fast, for example and in a simple manner, the number of eligible users are limited so that then complex analysis method for identification can be connected using the biometric data in a further step.

The further analysis of the biometric access sample All or part of the deposited for the users of the system Biometric samples can vary in a variety of ways Method of the prior art done. In particular, can the analysis of the biometric data to identify the user include a variety of statistical and mathematical methods. But even a simple feature comparison can be applied here.

By the procedure that the analysis of biometric data not only against a stored reference pattern of an alleged User takes place, but against several deposited samples For example, you can also analyze against biometric samples have been made by so-called "users" are those who have finally been denied access, for example to determine if maybe a repeated attack on the access control takes place.

One corresponding method can be realized by means of a software product If, for example, a data processing system with appropriate Acquisition devices for the biometric data is available, such as a keyboard of a PC on which key sequences can be entered, or Fingerprint scanner, eye scanner and the like. Corresponding a device according to the invention can be a data processing device with corresponding input means.

BRIEF DESCRIPTION OF THE FIGURES

Further Advantages, characteristics and features of the present invention in the following detailed description of embodiments clearly with reference to the attached drawings. The painting show here in a purely schematic way in

1 a one-dimensional vector space;

2 a three-dimensional vector space; and

3 a two-dimensional vector space with the assignment of an access feature vector to certain areas of the vector space.

EMBODIMENTS

The 1 shows a first embodiment of the application of the present invention for identifying a user of a system. A corresponding system may be, for example, access control to a building or factory premises or access control to a computer network, such as a company's intranet network.

Used according to the invention the system biometric data to identify the users of the system. For example, in the case of access control, a fingerprint sensor, an eye scanner, a face scanner or a keyboard or a Keypad or corresponding sensor surfaces for actuation with your fingers to enter a string such as a short text be provided.

The Users of this system must have at least one biometric Deliver sample, which is deposited as a reference in the system. This means for the different biometric methods, that, for example, at least one fingerprint of a finger, at least one eye scan of an eye or the photograph of the face or a key sequence entered must be stored.

at The access control will then be corresponding biometric data recorded, so again either taken a fingerprint, an eye scanned, the face imaged or entering a string required in a keyboard or on a keypad.

at the method according to the invention will now be apparent determines the deposited biometric samples feature vectors and saved. The feature vectors can be in a vector space be arranged so that for all biometric samples, which records the system, stores corresponding feature vectors are, which form a so-called biometric map.

An example of a one-dimensional vector space is in 1 shown. In the one-dimensional vector space-dimensional feature vectors are N _1, N _2, N ₃ and N ₄ deposited for the users 1, 2, 3 and 4. FIG. The one-dimensional feature vector N ₁ to N ₄ may be, for example, the average holding duration of a key in a tip sample. Accordingly, the holding period of the user 1 is smaller than the holding period of the users 2, 3 and 4.

In the access control, a corresponding access feature vector Z is now formed from the access sample which the user issues also in the vector space, for example the 1 , can be represented. From a comparison of the access feature vector Z with all the stored feature vectors N ₁ to N ₄ or a part of the feature vectors, for example the nearest feature vectors N ₁ and N ₂ , the user closest to the access feature vector can, for example, the feature vector N _{1 of} the user N _{1 are} identified. In order to obtain a certain degree of certainty, a minimum match, that is to say a maximum distance of the access vector Z from the feature vector of the user N _{1, may be} required. In this simplest case, only one feature of the biometric samples is compared and assigned, and this can be done without knowing the alleged identity of the user. Thus, no comparison is made of the access sample with only one biometric sample deposited, but it is compared with several stored biometric samples, so that the identification certainty is increased by the risk of accidental comparison with a known stored biometric sample Matching the biometric samples can be reduced or eliminated, since when matched with multiple feature vectors from multiple different users, the system can not make a unique association, thus requiring further verification. Thus, with this very simple and effective approach, at least the access control within a certain group of users can be effectively designed.

to Increase the security of identification either several vector spaces with different characteristics and / or multidimensional vector spaces.

An example of a three-dimensional vector space is in 2 shown. On the coordinate axes of the three-dimensional vector space, the features M ₁ , M ₂ and M ₃ are shown, such as again average holding time of a key, transition period from one key to the other key, frequency of the key operation, etc. In the vector space of 2 the feature vectors N ₂₁ , N ₂₂ , N ₂₃ , N ₁₁ , N ₁₂ , N ₁₃ , N ₃₁ , N ₃₂ , N ₃₃ are shown. Correspondingly, three biometric samples are dispensed by three users N ₁ to N ₃ . The vector N ₁₁ , ie the biometric sample 1 of the user 1, comprises the features M ₁₁₁ , M ₂₁₁ and M ₃₁₁ .

Again, similar to the embodiment of the 1 an access feature vector Z of an access sample is determined and displayed in the vector space, but this is to maintain the clarity of the diagram in the 2 has not been shown.

The 3 shows a two-dimensional vector space with the features M ₁ and M ₂ . The individual feature vectors of the users N ₁ and N ₂ are only represented by corresponding points. The feature vectors of the user N ₁ define an area in the vector space represented by a corresponding circle. Accordingly, users can be defined N ₂ , N ₃ and N ₄ areas. The definition of a corresponding user area can be defined by a corresponding envelope, in the case of two-dimensional vectors, for example, by an enclosing circle, in multidimensional vector spaces by a corresponding sphere or hyper sphere. In addition, the range for the individual user can be extended by a similarity range, so that, as represented, for example, for the user N ₂ , a typical range T ₂ results.

Accordingly, when determining an access feature vector Z, it can be determined whether the access feature vector Z lies in a user area, such as the user areas N ₃ and N ₄ , or in a typical area T ₂ . In the situation presented the 3 For the access sample, the three users N ₃ , N ₂ and N _{4 would be} considered. Accordingly, by preselecting or assigning the access feature vector to stored feature vectors, a preselection can be made as to which users are eligible for the identification. In the illustrated embodiment of the 3 the user N ₁ leaves because the user area of the user N ₁ lies outside the coordinate point of the access feature vector Z.

This can then be followed by a corresponding analysis of the biometric sample by conventional methods, whereby the effectiveness can be increased by the appropriate pre-selection of eligible users and in addition by analyzing the access sample against several eligible users, for example in the case of 3 It can be ruled out against the users N ₂ , N ₃ and N ₄ that a false identification can take place by accidental coincidences or by insufficient separation of the corresponding biometric samples. In the example of the 3 For example, if the access probe is only checked against the user N ₃ deposited biometric samples, then the user N _{2 could be} mistakenly admitted as having apparently biometric data similar to that of the user N ₃ . By knowing that the access sample is already in the similarity range of the user N ₂ and in the user area of the user N ₃ , an identification can be ruled out by a correspondingly more complex analysis of the biometric samples or analysis with narrower limit values.

Even though the present invention with reference to the embodiments has been described in detail, is for the expert of course, that the invention is not limited by these Embodiments is limited, but that rather modifications, for example, by omitting individual features or other combinations of individual vorgestell ter characteristics are possible without the scope of protection of the attached To leave claims. In particular, the present invention includes Invention all combinations of all features presented.

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• WO 98/06020 A2 [0002]
• - US 6151593 A [0002]

Claims (26)

A method for identifying a user of a system by means of a biometric sample having user-typical features, characterized in that for each user of a plurality of users of at least one biometric sample of the user at least one feature vector is stored in at least one vector space and that for the to be identified User is generated a corresponding access feature vector from an access probe, wherein for identification, a comparison of the access feature vector with at least one feature vector of at least one vector space. Method according to claim 1, characterized in that that the vector space is a one- or multi-dimensional space. Method according to one of the preceding claims, characterized in that a plurality of vector spaces are stored become. Method according to one of the preceding claims, characterized in that a feature vector is one or more Features of a biometric sample. Method according to one of the preceding claims, characterized in that the biometric sample is selected from the group is selected, the fingerprints, eye and Includes face pictures and typing behavior on a keyboard. Method according to claim 4 or 5, characterized that at least one element selected from the group as a feature which is the holding time of a key, the transition period by releasing a key to press the next key, the transition period from releasing a key to releasing the next key, the transition time from pressing a key to press the next key, the Error frequency or frequency, the Tipprate, the attack frequency, single or multiple overhauls, the frequency the use of certain function keys, the selection of alternatives Keys, selecting the left or right shift key, statistical Values and temporal or key-related abnormalities, characterized by extraordinary values from the highlight the data collected. Method according to one of the preceding claims, characterized in that feature vectors for each user several biometric samples are stored. Method according to one of the preceding claims, characterized in that in the vector space only feature vectors certain users or user groups are summarized. Method according to one of the preceding claims, characterized in that defines a region in the vector space is where the feature vectors of a user lie. Method according to one of the preceding claims, characterized in that in the vector space at least one typical Range defined for one or more feature vectors are. Method according to one of the preceding claims, characterized in that the identification without knowledge of alleged identity of the user. Method according to one of the preceding claims, characterized in that the comparison of the access feature vector with all feature vectors or only a part of the feature vectors takes place in a part of the vector space, which in a defined similarity range lies to the access feature vector. Method according to one of the preceding claims, characterized in that the comparison of the access feature vector only with feature vectors of a user group. Method according to one of the preceding claims, characterized in that the comparison is the determination of the Access feature vector nearest feature vector. Method according to one of the preceding claims, characterized in that the comparison is the assignment of the access feature vector one or more users and / or one or more typical ones Areas and / or neighboring areas to the typical areas. Method according to one of the preceding claims, characterized in that the comparison of the access feature vector and at least one feature vector a preselection for determination potential identities, which is a verification identity by analysis of the biometric access probe followed. Method according to one of the preceding claims, characterized in that an analysis of a biometric access probe towards all or part of it for the users the system deposited biometric samples. Method according to one of the preceding claims, characterized in that an Ana analysis of the biometric access sample only takes place with respect to those or those stored biometric samples of the users of the system, which have been determined when comparing the access feature vector with at least one feature vector. Method for identifying a user of a Systems by means of a biometric sample, which typical user Features, characterized in that the identification an analysis of at least one user's access sample all or part of the deposited for all users of the system biometric samples. Method according to claim 18, characterized that identification without knowledge of the alleged identity the user takes place. Method according to one of claims 15 to 20, characterized in that the analysis is biometric Samples are taken from at least two different users. Method according to one of claims 16 to 21, characterized in that the part of the user opposite whose biometric samples are analyzed by the access sample are selected by certain attributes, being biometric Samples of unsuccessful access attempts by the non-users group can be assigned. Method according to one of claims 15 to 21, characterized in that the analysis of the biometric data comprises at least one method from the group, which is a feature comparison of Access sample with edited or unprocessed, deposited Features of one or more users and a preparation of the access probe Includes statistical and mathematical methods. Method according to one of claims 15 to 23, characterized in that the analysis is the feature comparison with multiple biometric samples of one or more users. Software product with a stored command sequence for execution on a data processing device, wherein upon execution of the instruction sequence a method one of the preceding claims becomes. Apparatus for carrying out the method according to one of claims 1 to 22 with a data processing device and at least one input device for inputting a biometric Sample which is prepared such that a method according to of claims 1 to 22 is performed.