• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The invention relates to a method for enrolling biometric data in a database, each data comprising an information vector on a biometric trait, and a mask vector, determining the bits of the information vector to be taken into account for a comparison of data, the method comprising the application of a permutation on the bits of the vectors, the method being characterized in that it further comprises a step of encoding the vectors by an enrollment code, the permutation being implemented on the encoded vectors, and said encoding comprises: the representation of each bit of the mask vector in a sequence of several bits, such that the average weight of the representations of the set of bits of the mask vector is constant or statistically constant regardless of the bit values of the mask vector, and the representation of each bit of the information vector in a sequence comprising at least one bit ti randomly, the bits drawn randomly according to the same distribution law as the bits of the information vector. The invention also relates to a database, and a data acquisition method, comprising encoding said data for comparison with the data of the database.
    公开号:FR3018934A1
    申请号:FR1452444
    申请日:2014-03-24
    公开日:2015-09-25
    发明作者:Julien Bringer；Herve Chabanne
    申请人:Morpho SA；
    IPC主号:
    专利说明:

    [0001] FIELD OF THE INVENTION The field of the invention is that of data enrollment methods, particularly biometric data, in a database, guaranteeing the security of the data once enrolled in the database, and databases comprising such data. The invention applies in particular to the enrollment of biometric iris data. STATE OF THE ART Biometric data, in particular of iris or vein, generally comprises two identical-sized binary vectors: a first vector, termed information, contains the information on a biometric trait acquired on a person. The second vector, called mask, includes the information of whether corresponding parts of the information vector must be taken into account to perform a data comparison. For example, in the case where a biometric data is an iris image, the information vector is for example generated by acquiring a color image of the iris, and then converting this image into shades. of gray, and end by operating a threshold on the shades of gray to obtain binary information. The information vector is organized such that adjacent areas of an iris correspond to successive bits in the vector. On the other hand, the mask vector, which comprises a set of indexed bits corresponding to the bits of the information vector, determines the bits of the information vector to be taken into account for a comparison. If for example a part of an iris is obscured by an eyelid, regardless of the value of the bit in the information vector corresponding to this part, the bit of the corresponding mask vector indicates by its value, 0, or 1, do not take it into account.
    [0002] Such data are stored in databases for making identifications or authentications of individuals by comparison with other biometric data. The comparisons are most often made by calculating the Hamming distance between two data, which lists the number of different pixels between the two data.
    [0003] For data of the type described above, which include a mask vector, the Hamming distance is written: (ieir) n (mn m ') where i and i' are the information vectors of two data, and m and m 'are the mask vectors, which comprise bits at 1 for the unobstructed zones, or zones to be taken into account for the comparison, and bits at 0 for the zones to be ignored for comparison ( obscured or questionable areas). Thus, the Hamming distance is calculated by taking into account only those parts of the information vectors to be taken into account for the comparison.
    [0004] In order to protect such data stored in databases, it has been proposed to apply a permutation on the information vector and the mask vector, the permutation being identical for the two vectors. This makes it possible to preserve the Hamming distance between two data permutated identically.
    [0005] Now, it is possible to find the permutation used on a biometric datum from the pair x, p (x), where x denotes a datum comprising an information vector and a mask vector, and p (x) the permuted datum. , or even from a couple x, p (x ').
    [0006] Indeed, to find the permutation we can always exploit, the bits of the mask vector to 0 to find the geographical areas of the iris that are hidden (for example by eyelashes or eyelids). It is also possible to exploit the links existing between successive bits of the information vector, which correspond to geographical correlations in zones of the iris.
    [0007] Therefore, it is necessary to further protect the biometric data of the type comprising two vectors: an information vector, and a mask vector.
    [0008] PRESENTATION OF THE INVENTION An object of the invention is to provide a method of enrolling biometric data with increased security compared to the prior art.
    [0009] Another object of the invention is to propose a data enrollment method for a database, and a method for acquiring data to be compared with data from the database, which makes it possible to calculate a distance Hamming between the two data.
    [0010] In this regard, the subject of the invention is the method of enrolling biometric data in a database, each data item comprising a binary information vector on a biometric trait, and a binary mask vector determining the bits of the vector. of information to be taken into account for a data comparison, the method comprising applying a bit swapping of the information vector and the mask vector, the method being further characterized by step of encoding the vectors by a so-called enrollment code, the permutation being implemented on the encoded vectors, and said encoding comprises: the representation of each bit of the mask vector in a sequence of several bits, such that the mean weight of the representations of all the bits of the mask vector is constant or statistically constant regardless of the bit values of the mask vector, and - the representation of each bit of the information vector in a sequence comprising at least one randomly drawn bit, the bits drawn randomly according to the same distribution law as the bits of the information vector. Advantageously, but optionally, the enrollment method according to the invention may further comprise at least one of the following features: the sequence of bits representing the encoding of a bit of the information vector to be taken into account for a data comparison comprises at least one randomly drawn bit and the bit of the information vector. the permutation implemented on the vectors and encoded is identical, and the position of the bit of the information vector in the bit sequence representing the encoding of a bit to be taken into account for a data comparison corresponds to the position a bit to 1 of the bit sequence representing the encoding of a bit of the mask vector. the position of the bit of the information vector in the bit sequence representing the encoding of a bit to be taken into account for a data comparison does not correspond to the position of a bit at 1 of the sequence of bits representing the encoding of a bit of the mask vector, and the permutation step comprises the application of distinct permutations on the bit sequences representing the encodings of the information vector and the mask vector, the permutations being adapted so that the position of the bit of the information vector after permutation corresponds to a bit at 1 of the bit sequence of the encoded mask vector after permutation. - The sequence of bits representing the encoding of a bit of the information vector not to be taken into account for a data comparison comprises a single randomly drawn bit or a randomly drawn bit and the bit of the information vector. the sequence of bits representing the encoding of a bit (m) of the mask vector comprises a sequence Y of n-1 bits and the bit of the mask vector, the sequence Y of bits being able to take 2n-1 values Y1, ..., Y2A (n-1), and the probabilities al, ... a2. (N_1) to choose the value Y1, ..., Y2. (N_1) when the bit of the mask vector is 0 and b1 , ..., b2. (n_1) to choose the value Y1, ..., Y2A (n-1) when the bit of the mask vector is 1 is as follows: [Weight (Yi) * ai] = I [ (Weight (Yi) + 1) * i = t - the values that can take the sequence Y of bits are set or pulled randomly. the number of bits at 1 of a sequence of bits representing the encoding of a bit of the mask vector has a statistically constant average weight of n / 2. - The method further comprises, after the application of the permutation, the multiplication of the mask vector encoded by an invertible matrix. The subject of the invention is also a database comprising at least one biometric data enrolled by the implementation of the enrollment method according to the preceding presentation. Another object of the invention is a method of acquiring data. a new biometric data item for its comparison with a datum enrolled in a database according to the preceding presentation, including the application of a permutation to the new biometric data, characterized in that it furthermore comprises the encoding of the new biometric data by a verification code, the permutation being implemented on the encoded data, said encoding and permutation being adapted to preserve the value of the Hamming distance or the weighted Hamming distance between the new data and the data base after their respective encoding.
    [0011] Advantageously, but optionally, during the acquisition process of a new datum, the new datum may comprise a binary vector of information on a biometric line, and a binary vector of a mask determining the bits of the information vector to be taken into account. for the comparison, and the encoding of the new data comprises the representation of each bit of the data vectors by a sequence comprising a number of bits equal to the number of bits in the sequences representing the encoded vectors of the data of the database, the permutation applied to the new data being identical to the permutation applied to the data of the database during its enrollment, and the representation of the bits of the mask vector being further adapted so that only the intersection of the two bit sequences representing the bits mask vectors corresponding to areas to be taken into account for the comparison is non-zero.
    [0012] According to another aspect, the subject of the invention is a method for comparing a biometric data item enrolled in a database according to the preceding presentation, and a biometric data item acquired according to the acquisition method, the method comprising: generation of an inverse matrix of the invertible matrix multiplied by the vector of the mask during the enlistment of the datum of the database, - the multiplication of the sequence of bits representing the vector of the encoded mask of the datum of the database by the inverse matrix and the calculation of a weighted Hamming distance between the new data encoded by the verification code and the encoded data of the base obtained at the end of the multiplication.
    [0013] The invention finally relates to a system comprising a database according to the preceding presentation, and at least one database management server, comprising processing means adapted to implement, on a data comprising two vectors. binary, an enrollment encoding method, comprising: - representing each bit of the mask vector in a sequence of several bits, such that the average weight of the representations of all the bits of the mask vector is constant which the values of the bits of the mask vector, and - the representation of each bit of the information vector in a sequence comprising at least one randomly drawn bit, the randomly drawn bits according to the same distribution law as the bits of the vector d 'information. The proposed enrollment method makes it possible to increase the security of data enrolled in a database, since the encoding makes it possible to balance the distribution of the bits of a mask vector, so as not to betray, by the position of the data elements. 0 or 1, the position of a characteristic point of the biometric data such as the eyelid for an iris image. The enrollment encoding also makes it possible to eliminate the links between successive bits of the information vector by adding a random component in the encoded representation of this vector. Keeping the same distribution law as the information vector makes the random components undetectable. The proposed acquisition method also makes it possible to compare a new datum with a datum enrolled in the database by the method described above by calculating a Hamming distance or weighted Hamming distance between these data, by adapting the data. verification encoding to the enrollment encoding. DESCRIPTION OF THE FIGURES Other characteristics, objects and advantages of the present invention will appear on reading the detailed description which follows, with reference to the appended figures, given by way of non-limiting examples and in which: FIG. schematically a system comprising a database; - Figure 2 represents the main steps of a method of enrolling data in a database. FIG. 3 represents the main steps of a method of acquiring a new datum and of comparing it with datum of the database.
    [0014] DETAILED DESCRIPTION OF AT LEAST ONE EMBODIMENT Data storage system With reference to FIG. 1, there is shown a system 1 comprising a database 10 in which digital data in the form of data sequences are recorded or enrolled. bits, also called in the following bit vectors. The data are advantageously biometric data, such as, for example, images of iris or venous networks.
    [0015] Each datum comprises two binary vectors: a first vector I, called information, containing the information on a biometric trait acquired on a person, and the second vector M, called a mask, including information on whether corresponding parts of the information vector must be taken into account to perform a data comparison.
    [0016] According to a first embodiment, the two bit vectors I and M are of identical sizes. Thus, each vector I, M comprises a sequence of bits resp. i, m, indexed, and the value of a bit m of the mask vector M indicates whether the bit i of the same index of the information vector I must be taken into account to perform a data comparison.
    [0017] Alternatively, the two I and M vectors may be of different sizes. For example, a geographical position of a biometric feature may be represented by several bits i of the information vector I and a single mask bit m. In the following example, it is considered that a bit at 1 of the mask vector M indicates one or more bits to take into account the information vector I, and a bit at 0 indicates one or more bits to be ignored. in account of the information vector I. In the case of an iris image, it may be an area obscured by an eyelid or eyelashes. The system 1 further comprises a database management server 11, adapted to access the database 10 for reading and writing, and which comprises processing means, for example a processor, making it possible to implement the enlistment described below, as well as, where appropriate, the acquisition of a new datum and the comparison of this datum with data from the database.
    [0018] In the case where the database 10 stores biometric data, this database 10 is therefore used to perform authentication or biometric identifications on individuals from biometric data acquired on the individual and compared to the data of the database. The system 1 further comprises a biometric data sensor 12, which is chosen according to the nature of the biometric feature on which the data is taken. This sensor 12 may be used to acquire data to be enrolled in the database 10 or to acquire new data to be compared to data from the database during authentication or identification.
    [0019] Presentation of the enrollment method A method of enrolling 100 of a datum in the base 10 will now be described with reference to FIG. 2. A first step 110 of this method consists in obtaining data to be enrolled in the database. 10. This step can be implemented by capturing data on an individual by the sensor 12, or by recovering the data on another medium. For example, this data could be in a network to which the management server 11 has access, or in an identity document of an individual, etc. The data thus recovered comprises the two bit vectors I and M described above. This data is then encoded 120 by a first enrollment code enc_enrol. This step is preferably implemented by the management server 11 and is described in more detail below.
    [0020] As described below, the method then comprises the application of a permutation 130, adapted to the encoding, preferably identical to the encoded representations of the information and mask vectors. Optionally, the management server may further multiply the encoded and permuted mask vector with a randomly VV invertible matrix during a step 140, thereby making the transformation to the vectors more complex and it is therefore more difficult to find the starting information. Finally, during a step 150, the management server 11 records in the database the data thus encoded.
    [0021] Enrollment Encoding Back to the enrollment encoding step 120, this encoding is applied to the two bit vectors I, M in different ways. The enc_enrol enrollment encoding applied to the mask vector M comprises the representation of each bit m of the vector by a sequence of bits comprising at least two bits, the bit sequences representing the encoded bits of the mask vector having a constant average weight or statistically constant over the entire encoding of the mask vector, regardless of the bit values of the mask vector.
    [0022] By statistically constant average weight, it is meant that the average weight of the bit sequences over the entire encoding of the mask vector M tends to a fixed value. Advantageously, the sequence of bits representing the encoding of a bit m of the mask vector comprises n bits, of which the first n-1 form a vector Y = and the last of which is the bit of the mask vector. The vector Y can be fixed in a determined manner, by associating a value for m = 0, and another value for m = 1, or randomly drawn, but in such a way as to respect the constraint that the number of bits at 1 (and therefore to 0) in YI Im is constant on average on the encoded representation of the mask vector M, that m is equal to 0 or 1. We can satisfy this condition, if we write Y1 ... 1 .... Y 2 ^ (n-1) the possible values of the vector Y, by associating a probability a, taking the value Y, for m = 0, and a probability [3, taking this value for m = 1 , the probabilities a, and [3, being constrained as follows: 2n-1 2n-1 1 [Weight (Y) * ai] = I [(Weight ((Yi)) + 1) * i = t The weight of a sequence of bits is the Hamming weight, that is, the number of bits at 1 in the sequence. This condition is equivalent to the other following condition (which is therefore satisfied if the first condition is): [n - Weight (Yi) * ai] = 1 [n - 1 - Weight ((Yi)) * i = t The proposed encoding makes it possible to render non-distinguishable a "masked" position, that is to say corresponding to a bit of the information vector not to be taken into account for the comparison, of an "unmasked" position. looking at a bit in isolation. Indeed, when the mask vector is not encoded, it is sufficient to examine the values of the bits to obtain information on the areas to be taken into account or not for a comparison between two data. However, once the encoded mask vector and the permutation applied to the bits of the representation of the vector, it is not possible to say, considering a single bit, whether this bit comes from the representation of a bit at 1 or from one bit to 0 of the mask vector. For example, the encoding of the mask vector can be performed as follows: Mask bit = 0 (Y = 1) Mask bit = 1 01 (Y = 0) Alternatively, when the mask vector bits are encoded as sequences of three bits, the encoding may comprise the representation of a bit at 0 in 110, 100, 010, or 000 and respectively the representation of a bit at 1 in 111, 101, 011, or 001.
    [0023] We thus have: Y1 = 11, Y2 = 10, Y3 = 01 and Y4 = 00. We can for example choose the probabilities a, and [3, as follows: = = 5/8, and a2- 3- a4- [ 32-33-1 / 8. The larger the number of bits in a Y sequence, the more secure the data encoding.
    [0024] An additional condition for further increasing the protection of the data encoded in the base is to set, on the set of encoded representations of the bits of the mask vector, the weight of the representations YI lm on average at n / 2 (where n is the length of the bit sequence representing the encoding of a bit), whether m is equal to 0 or 1. According to the preceding example, the statistically constant average weight equal to n / 2 is obtained by adjusting the probabilities a, and [3, so that the average weight of the data is statistically equal to n / 2.
    [0025] Enc_enrol enrollment encoding applied to the information vector comprises the representation of a bit of the vector by a sequence of bits comprising at least one randomly drawn bit, but so that the randomly drawn bits follow the same law. of distribution as the information vector.
    [0026] This makes it possible to eliminate the correlations between successive bits of the information vector, to eliminate the possibility of extracting information on the encoded data. Advantageously, the encoding of bits to take into account the information vector, that is to say according to the example described above, bits corresponding to bits of 1 of the mask vector, advantageously comprises at least one bit randomly drawn and the initial bit of the information vector. This makes it possible to preserve the information contained in the information vector in order subsequently to implement a comparison between two data, by calculating a Hamming distance or a weighted Hamming distance, according to an advantageous embodiment example. of the invention, or another comparison function commonly used in the field. Advantageously, the position of the initial bit of the information vector in the bit sequence corresponds to the position of a bit at 1 of the bit sequence representing the encoding of the bit of the corresponding mask vector. This makes it possible to preserve the information contained in the vector for the calculation of the Hamming distance, and thus to preserve this distance during the encoding of the data. However, a bit of the information vector corresponding to a masked area (therefore in the example to a bit at 0 of the mask vector) may not include the initial bit of the vector, and be limited to one or more randomly drawn bits. . Alternatively, in order not to make any distinction in the coding of the bits of the information vector, the encoded representation of a bit not to be taken into account for a comparison still includes the initial bit as well as one or more bits. randomly drawn. For example, the preferred embodiment of the information vector enrollment encoding is adapted to the previous encoding of the mask vector and includes representing each bit of the information vector with a sequence of Ahi bits. where A = aill ... 11a (, 4) and i is the bit of the information vector. The a, are bits drawn at random but respecting the distribution of the bits i of the information vector. For example, if the probability P (x = 0) = h, the ai are drawn with the same probability of being equal to 0. To repeat the example given previously on two-bit sequences representing the binary vectors, We obtain the following encoding: Enc_enrol (x, 0) = (ax, 10) Enc_enrol (x, 1) = (a'x, 01). Where a and a 'are bits drawn randomly and following the distribution of the bits of the information vector.
    [0027] To also resume the previous example given on sequences of three bits representing the binary vectors, we also obtain the following encoding: Enc_enrol (x, 0) = (abx, 110) or (abx, 100) or (abx, 010 ) or (abx, 000) Enc_enrol (x, 1) = (cdx, 111) or (cdx, 101) or (cdx, 011) or (cdx, 001).
    [0028] Here again, the implementation of the permutation on the encoded representation of the bits of the information vector makes it impossible to obtain, by considering a bit in isolation, information on the initial bit. According to an alternative embodiment, the position of the initial bit of the information vector in the bit sequence does not correspond to the position of a bit at 1 of the bit sequence representing the bit encoding of the corresponding mask vector. , but then the permutations applied to the sequences of encoded bits of the I and M vectors during step 130 are different, and are adapted so that the position of a bit of the information vector, after application of the permutation, corresponds to the position of a bit at 1 after permutation of the bit sequence representing the encoding of the mask vector. This generally applies whether the bit of the information vector is a bit to be taken into account (mask bit at 1) or not (mask bit at 0).
    [0029] This variant also makes it possible to keep the value of the Hamming distance between the data. With reference to FIG. 3, a method of acquiring a new datum 200, preferably a biometric datum, will now be described in order to compare it with a datum enrolled in the database 10 according to the preceding method. This method is adapted to the above enrollment method to allow a Hamming distance (including a weighted Hamming distance) to be calculated on the encoded data, and thus to preserve the result of this distance between the two before and after data. encoding. To do this, the method 200 comprises a first step 210 of acquiring a new datum. This step is advantageously implemented by the sensor 12 by acquisition of a biometric trait on an individual, which then transfers the data to the management server 11. Alternatively, this step 20 is performed by recovering a biometric data on a network or on an identity document. The data obtained comprises, as the data processed by the enrollment method, two vectors of identical size; an information vector and a mask vector. The size of these vectors is also identical to the size of the vectors of the data processed by the enrollment method 100. The method 200 then comprises a step 220 of encoding the data thus obtained by a verification code called audit, which may be different from the enrollment code, and which is adapted to it to preserve the Hamming distance between the data. In this regard, each encoded representation of a bit of the information vector and the mask vector comprises a number of bits identical to the bit sequences respectively representing the encoded information and mask vectors of the enrolled data item.
    [0030] In addition, the encoding of the bits of the mask vector is adapted so that only the intersection of bits corresponding to areas of the information vectors to be taken into account for a comparison is non-zero. Indeed, the weighted Hamming distance is written: 11 (1 st) n According to the above example, in which the bits at 1 of the mask vectors m, m 'correspond to unobstructed areas of a biometric feature , and therefore to take into account for a comparison, and in which the encoding of enrollment of a bit of the mask is written Yllm = the encoding of a bit m 'of the mask vector of the new data is advantageously 011 ... 11011m.
    [0031] To resume the above example where the enrollment encoding is implemented on two bits, the verification encoding enc_verified is therefore as follows: Mask bit = 0 00 Mask bit = 1 01 So we need to have the two mask bits initial to 1 to have a non-zero intersection: ionoo = oo; ionol = oo, oonol = oo; Oinoi = oi. Regarding the verification encoding (enc_vérif) of the information vector, it suffices that the sequences obtained have the same number of bits as the sequences encoded by the enrollment code, and that the representation of a bit corresponding to a zone to take into account for the comparison comprises said bit, at a position opposite a bit to 1 of the encoded representation of the corresponding bit of the mask vector. According to the preceding example, where the enrollment encoding of an information bit is written as Ali, the encoding of an information bit i 'of the new datum may be of the type where A' = a 'i 11 ... 11a' (, 4) and the, are randomly drawn. We thus obtain the following encoding, according to the preceding example: Enc_verify (x, 0) = (ax, 00) Enc_enrol (x, 1) = (bx, 01).
    [0032] The implementation of the permutation on the bit sequences representing the bits of the information vector permits this to separate the bits of the bit sequences representing the same initial bit of the information vector and the mask vector. It is therefore impossible, by examining a bit in isolation, to deduce the initial value of the bit of which this bit forms part of the encoded sequence.
    [0033] The method 200 then comprises the application 220, by the management server 11, on the encoded representations of the information and mask vectors, of the same permutation as that applied to the enrolled data. This makes it possible to subsequently implement a comparison 300 between the acquired data item and a data item of the base by calculating the Hamming distance on the encoded data. In this respect, the management server 11 of the base retrieves 310 data stored in the base in encoded and permuted form. Optionally, in the case where the enrollment of the data in the database 10 comprises the multiplication of the mask vector encoded and permuted by an invertible matrix, the method 200 further comprises the calculation 320 of the inverse matrix of the invertible matrix, and multiplying the bit sequence recorded in the base, corresponding to the encoded and permuted mask vector, by the inverse matrix.
    [0034] Finally, the management server 11 performs the comparison 330 between the two data by calculating the weighted Hamming distance between them, according to the formula provided above. The proposed method therefore makes it possible, by encoding the mask vector followed by the permutation, to make the masked areas of the non-masked areas non-distinguishable by considering the bits in isolation, and by encoding the information vector, remove the links between two successive bits of the information vector. In addition, the enrollment encoding and the check encoding preserve the value of the Hamming distance during encoding, which makes the execution of the comparison faster.
    权利要求:
    Claims (13)
    [0001]
    REVENDICATIONS1. Method for enrolling (100) biometric data (I, M) into a database (10), each data comprising an information bit vector (I) on a biometric trait, and a mask bit vector (M) determining the bits (i) of the information vector to be taken into account for a data comparison, the method comprising applying (130) a bit permutation of the information vector (I) and the vector of mask (M), the method being characterized in that it further comprises a step of encoding (120) the vectors by a so-called enrollment code (enc_enrol), the permutation being implemented on the encoded vectors, and said encoding comprises: - representing each bit (m) of the mask vector (M) in a sequence of several bits, such that the average weight of the representations of the set of bits of the mask vector (M) is constant or statistically constant regardless of the bit values (m) of the mask vector and the representation of each bit (i) of the information vector (I) in a sequence comprising at least one randomly drawn bit (a), the randomly drawn bits following the same distribution law as the bits of the vector of information.
    [0002]
    The method of enrollment (100) according to claim 1, wherein the bit sequence representing the encoding of a bit (i) of the information vector (I) to be taken into account for a data comparison comprises minus one bit randomly drawn (a) and the bit of the information vector (i).
    [0003]
    The method of enrollment (100) according to claim 2, wherein the permutation set up on the vectors (I) and (M) encoded is identical, and the position of the bit (i) of the information vector (I ) in the bit sequence representing the encoding of a bit to be taken into account for a data comparison corresponds to the position of a bit at 1 of the bit sequence representing the encoding of a bit of the mask vector (M).
    [0004]
    The enrollment method (100) according to claim 2, wherein the bit position of the information vector (I) in the bit sequence represents the encoding of a bit to be taken into account for a data comparison. does not correspond to the position of a bit at 1 of the bit sequence representing the encoding of a bit of the mask vector, and the permutation step (130) comprises the application of distinct permutations on the sequences of bits representing the encodings of the information vector and the mask vector, the permutations being adapted so that the position of the bit of the information vector after permutation corresponds to a bit at 1 of the bit sequence of the encoded mask vector after permutation .
    [0005]
    5. Enrollment method (100) according to one of claims 1 to 4, wherein the bit sequence representing the encoding of a bit (i) of the information vector (I) not to be taken into account. for a data comparison comprises a single randomly drawn bit or a randomly drawn bit and the bit of the information vector.
    [0006]
    The enrollment method (100) according to one of the preceding claims, wherein the bit sequence representing the encoding of a bit (m) of the mask vector (M) comprises a Y sequence of n-1. bits and the bit (m) of the mask vector (M), the sequence Y of bits can take 21 values Y - 1, ---, Y2 ^ (ni), and the probabilities cub - a2A (n_l) to choose the value Y - 1, - - -, Y2 ^ (n-1) when the bit of the mask vector is 0 and 13-1, ..., b2. (, 1) to choose the value Y1 - - -, Y2A (ni) when the bit of the mask vector is 1 is as follows: 2n-1 {Weight (Y1) * ai] = [(Weight (Yi) + 1) * f31] [0007]
    7. Enlistment method according to the preceding claim, wherein the values (Y1) that can take the Y sequence of bits are set or pulled randomly.
    [0008]
    The enrolling method according to one of claims 6 or 7, wherein the number of bits at 1 of a bit sequence (Yll m) representing the encoding of a bit (m) of the present mask vector. a statistically constant average weight of 30 n / 2. 2n-1
    [0009]
    The enrollment method (100) according to one of the preceding claims, further comprising, after applying the permutation, multiplying (140) the mask vector encoded by an invertible matrix (W).
    [0010]
    10. Method of acquiring (200) a new biometric data item (I ', M') for its comparison with a datum enrolled in a database (10) by implementing the method according to one of the claims preceding, comprising the application of a permutation (230) to the new biometric data, characterized in that it furthermore comprises the encoding (220) of the new biometric data by a so-called verification code (enc_vérif), the permutation being implemented on the encoded data (enc_vérif (I ', M')), said encoding and the permutation being adapted to preserve the value of the Hamming distance or the weighted Hamming distance between the new data (I ', M') and the data base (I, M) after their respective encoding.
    [0011]
    11. Acquisition method (200) according to the preceding claim, wherein the new datum (I ', M') comprises an information binary vector (I ') on a biometric line, and a mask bit vector ( M ') determining the bits of the information vector to be taken into account for the comparison, and the encoding (220) of the new data item comprises the representation of each bit of the data vectors by a sequence comprising a number of bits equal to the number of bits in the sequences representing the encoded vectors of the data (I, M) of the base (10), the permutation applied to the new datum being identical to the permutation applied to the datum of the base (10) during its enrollment, and the representation of the bits of the mask vector is further adapted so that only the intersection of the two bit sequences representing the mask vector bits corresponding to zones to be taken into account for the comparison is non-zero the.
    [0012]
    12. A method of comparing (300) a biometric data item enrolled in a database (10) by implementing the method of claim 9, and an acquired biometric data item according to one of claims 10 or 11, the method comprising: - generating an inverse matrix of the invertible matrix multiplied by the vector of the mask upon enrollment of the data of the base, - multiplying (320) the sequence of bits representing the vector of the encoded mask of the data from the base by the inverse matrix, and the calculation of a weighted Hamming distance (330) between the new data encoded by the verification code and the encoded data of the base obtained at the end of the multiplication.
    [0013]
    13. System (1) comprising a database (10), and at least one management server (11) of the database (10), comprising processing means adapted to implement, on a data (I , M) comprising two bit vectors, a method (120) for enrolling encoding (enc_enrol), comprising: - representing each bit of the mask vector (M) in a sequence of several bits, such as the average weight representations of all the bits of the mask vector (M) are constant regardless of the bit values of the mask vector, and the representation of each bit of the information vector (I) in a sequence comprising at least one bit randomly drawn, the bits drawn randomly according to the same distribution law as the bits of the information vector.
    类似技术:
    公开号 | 公开日 | 专利标题
    EP2924609B1|2016-12-28|Method for enrolment of data in a database for the protection of said data
    EP2795831B1|2016-03-09|Biometric identification using filtering and secure multi party computation
    EP1811422B1|2016-08-31|Processes for determining an identifier, biometric verification and associated systems.
    EP2171913B1|2013-10-16|Processing of biometric data by transformation
    EP3206192A1|2017-08-16|Method for securing and verifying a document
    EP2494491B1|2017-07-12|Identification by means of checking a user's biometric data
    CA2743954C|2018-08-21|Identification or authorisation method, and associated system and secure module
    EP2705503B1|2016-06-22|Methods for biometric registration and verification, and related systems and devices
    EP2973210B1|2019-12-04|Secure data processing method, and use in biometry
    EP2862309B1|2016-03-02|Method of secure data processing
    EP2826200B1|2016-05-11|Method for encrypting a plurality of data in a secure set
    EP3043511B1|2017-11-08|Method for identifying an entity
    WO2009083528A1|2009-07-09|Method and system for generating stable biometric data
    EP3200387B1|2020-06-10|Secure multi-party processing method protected against a malicious party
    EP3825915A1|2021-05-26|Classification of a biometric print wherein an image is input
    EP3742699A1|2020-11-25|Method for strong authentication of an individual
    FR2962569A1|2012-01-13|METHODS, SYSTEMS, AND DEVICES FOR BIOMETRIC VERIFICATION
    FR2998391A1|2014-05-23|Method for identification and/or authentication of individual, involves implementing identification and/or authentication process such that image in comparison zone is compared with image in positioning zone after registration of images
    同族专利:
    公开号 | 公开日
    US20150269394A1|2015-09-24|
    JP6650204B2|2020-02-19|
    US9710631B2|2017-07-18|
    EP2924609A1|2015-09-30|
    KR20150110429A|2015-10-02|
    FR3018934B1|2017-05-26|
    JP2015207279A|2015-11-19|
    EP2924609B1|2016-12-28|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    CA2327911A1|2000-12-08|2002-06-08|Cloakware Corporation|Obscuring functions in computer software|
    US7237115B1|2001-09-26|2007-06-26|Sandia Corporation|Authenticating concealed private data while maintaining concealment|
    US7773784B2|2004-10-15|2010-08-10|University Of Colorado Board Of Regents|Revocable biometrics with robust distance metrics|
    JP5271669B2|2008-10-31|2013-08-21|株式会社日立製作所|Biometric authentication method and system|
    US20110047377A1|2009-08-19|2011-02-24|Harris Corporation|Secure digital communications via biometric key generation|
    WO2011113478A1|2010-03-16|2011-09-22|Carlo Trugenberger|Authentication system, method for authenticating an object, apparatus for producing an identication device, method for producing an identification device|
    US9621533B2|2012-12-14|2017-04-11|Nec Corporation|Bit string collation system, bit string collation method, and program|US10733415B1|2015-06-08|2020-08-04|Cross Match Technologies, Inc.|Transformed representation for fingerprint data with high recognition accuracy|
    CN105760818B|2016-01-28|2019-10-08|努比亚技术有限公司|A kind of eyeprint ciphering and deciphering device and method|
    CN106845265B|2016-12-01|2020-06-12|北京计算机技术及应用研究所|Document security level automatic identification method|
    US10873447B2|2018-05-24|2020-12-22|Visa International Service Association|Efficient concurrent scalar product calculation|
    JP2021525386A|2018-05-24|2021-09-24|ビザ・インターナショナル・サービス・アソシエーション|Efficient simultaneous scalar product calculation|
    US10713544B2|2018-09-14|2020-07-14|International Business Machines Corporation|Identification and/or verification by a consensus network using sparse parametric representations of biometric images|
    法律状态:
    2016-02-19| PLFP| Fee payment|Year of fee payment: 3 |
    2017-02-21| PLFP| Fee payment|Year of fee payment: 4 |
    2018-02-20| PLFP| Fee payment|Year of fee payment: 5 |
    2020-02-20| PLFP| Fee payment|Year of fee payment: 7 |
    2021-02-18| PLFP| Fee payment|Year of fee payment: 8 |
    2022-02-21| PLFP| Fee payment|Year of fee payment: 9 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1452444A|FR3018934B1|2014-03-24|2014-03-24|METHOD OF INPUTTING DATA IN A BASE FOR THE PROTECTION OF THESE DATA|FR1452444A| FR3018934B1|2014-03-24|2014-03-24|METHOD OF INPUTTING DATA IN A BASE FOR THE PROTECTION OF THESE DATA|
    EP15151412.2A| EP2924609B1|2014-03-24|2015-01-16|Method for enrolment of data in a database for the protection of said data|
    JP2015055623A| JP6650204B2|2014-03-24|2015-03-19|Method for registering data in the base and protecting the data|
    US14/665,955| US9710631B2|2014-03-24|2015-03-23|Method for enrolling data in a base to protect said data|
    KR1020150040921A| KR20150110429A|2014-03-24|2015-03-24|Method for enrolling data in a base to protect said data|
    [返回顶部]