System and method for biometric template security with password
A computer-implemented method is described for verifying a user with stored biometric templates. In an enrollment process, at least one biometric image of the user is provided along with a user-specified password. The biometric verification process uses the user-specified password only as needed, such as if the verification process has only a weak match. The method reduces the occurrence of false rejections that may occur for a variety of reasons. The method may also be extended and applied to a variety of types of secure image matching applications.
1 . A computer-implemented method, comprising:
receiving one or more first biometric images comprising at least one biometric characteristic of an individual for enrollment and verification;
identifying a plurality of interest points based on the one or more first biometric images and descriptors for the region around each interest point to generate an initial template;
generating obfuscating data points;
generating a biometric enrollment template having both interest data points and obfuscating data points;
encoding a secret key in a first augmented biometric enrollment template, with the secret key being reconstructible in a first verification process corresponding to a strong match to the plurality of interest points;
encoding the secret key in a second augmented biometric enrollment template with the secret key being reconstructible in a second verification process in response to the combination of a user-specified password and a weak match to the plurality of interest points, wherein the second augmented biometric enrollment template is encrypted by the user-specified password to generate an encrypted version of the second augmented biometric enrollment template to permit reconstructing the secret key in response to a combination of the weak match and the user-specified password; and
wherein in a two-stage authentication of a user from a received biometric image, the first verification process authenticates the user in response to detecting the strong match for the biometric image and wherein if the strong match is not satisfied the second verification process determines if the weak match is satisfied for the biometric image, and if the weak match is satisfied, the user is requested to input their user-specified password, with the second verification process authenticating the user in response to combination of the weak match being satisfied and the input user password matching the user-specified password.
2 . The computer-implemented method of claim 1 , wherein:
the first verification process requires at least k1 original points be identified for the strong match; and
the second verification process requires at least k2 original points be identified for the weak match if the user-specified password is also provided, where k1 and k2 are positive integers and k2 is less than k1.
3 . The computer-implemented method of claim 1 , wherein the weak match occurs for a low-quality sample image of the user.
4 . The computer-implemented method of claim 1 , wherein the weak match occurs for a new presentation of the user.
5 . The computer implemented method of claim 1 , wherein the weak match has a smaller number of matched image points than the strong match and the additional constraint of matching the user-specified password results in an entropy level being comparable to the strong match.
6 . The computer-implemented method of claim 1 , further comprising tagging one or more of the obfuscating data points using one or more of the interest points, wherein the tagging comprises providing as input to a tagging function one of the interest points and, in response to the input, receiving as output from the tagging function one of the obfuscating data points to be tagged.
7 . The computer-implemented method of claim 1 , wherein the one or more first biometric images comprise facial images, and wherein the interest points comprise facial regions.
8 . The computer-implemented method of claim 1 , wherein the one or more first biometric images comprise periocular images, and wherein the interest points comprise periocular regions.
9 . The computer-implemented method of claim 1 , wherein the one or more first biometric images comprise images of an eyeball, and wherein the interest points comprise vascular regions.
10 . The computer-implemented method of claim 1 , wherein the one or more first biometric images comprise images of an eyeball, and wherein the interest points comprise iridial regions.
11 . The computer-implemented method of claim 1 , further comprising associating one or more real descriptors with each interest point, wherein each real descriptor describes one or more localities surrounding the corresponding interest point.
12 . The computer-implemented method of claim 11 , further comprising associating one or more synthesized descriptors with each obfuscating data point, wherein each synthesized descriptor comprises a statistical similarity to the real descriptors.
13 . A computer-implemented method, comprising:
in an enrollment phase:
receiving a user-specified password;
receiving one or more first images comprising at least one biometric characteristic of the user for enrollment and verification;
identifying a plurality of interest points based on the one or more first images and descriptors for the region around each interest point to generate an initial template;
generating obfuscating data points;
generating an augmented template having both interest data points and obfuscating data points;
encoding a secret key in a first version of the augmented template that is reconstructible in a successful verification process requiring a strong match corresponding to at least k1 original points; and
encoding the secret key in a second version of the augmented template that is encrypted by a user specified password that is reconstructible in response to receiving the user-specified password from the user in combination with a successful verification process requiring a weak match to at least k2 original points, where k2 is less than k1, wherein the additional constraint of matching the user-specified password results in an entropy level comparable to the strong match condition; and
in a verification phase:
receiving a biometric image of a user;
determining whether the strong match exists to the augmented template matching at least k1 image points;
if the strong match exists, authenticating the user;
if the strong match doesn't exist, determining if the weak match exists having at least k2 image points, requesting the user provide the user-specified password, and determining if the weak match is satisfied and the user-specified password matches.
14 . The computer-implemented method of claim 13 further comprising tagging one or more of the obfuscating data points using one or more of the interest points, wherein the tagging comprises providing as input to a tagging function of one of the interest points and, in response to the input, receiving as output from the tagging function one of the obfuscating data points to be tagged.
15 . The computer-implemented method of claim 13 , wherein the one or more first images comprise images of a region of an eye, each eye region image comprising a view of a vasculature of the respective eye region, and wherein the interest points comprise vascular interest points.
16 . The computer-implemented method of claim 13 , further comprising associating one or more real descriptors with each interest point, wherein each real descriptor describes one or more localities surrounding the corresponding interest point.
17 . The computer-implemented method of claim 16 , further comprising associating one or more synthesized descriptors with each obfuscating data point, wherein each synthesized descriptor comprises a statistical similarity to the real descriptors.
18 . A system, comprising:
one or more computers programmed to perform operations including:
receiving one or more first biometric images comprising at least one biometric characteristic of an individual for enrollment and verification;
identifying a plurality of interest points based on the one or more first biometric images and descriptors a the region around each interest point to generate an initial template;
generating obfuscating data points:
generating a biometric enrollment template having both interest data points and obfuscating data points;
encoding a secret key in a first augmented biometric enrollment template, with the secret key being reconstructible in a first verification process corresponding to a strong match to plurality of interest points;
encoding the secret key in a second augmented biometric enrollment template with the secret key being reconstructible in a second verification process in response to the combination of a user-specified password and a weak match to plurality of interest points, wherein the second augmented biometric enrollment template is encrypted by the user-specified password to generate an encrypted version of the second augmented biometric enrollment template to permit reconstructing the secret key in response to a combination of a weak match and a user-specified password; and
wherein a two-stage verification process authenticates a user from a received biometric image in which the first verification process verifies a user in response to the strong match for the biometric image and if the strong match is not satisfied transitions to the second verification process in which a determination is made if the weak match is satisfied, the user is requested to provide their user-specified password, with the second verification process authenticating the user in response the user-specified password being received and there is the weak match to the biometric image.