Secure protection of biometric templates
This invention relates to methods and devices for verifying the identity of a person based on a sequence of feature components extracted from a biometric sample. Thereafter, the feature components are quantized and assigned a data bit sequence in such a way that adjacent quantization intervals have a Hamming distance of 1. The data bit sequences are concatenated into a bit string, and said bit string is combined with a helper data set by using an exclusive disjunction (XOR) operation into a codeword. Finally, the codeword is decoded into a secret V and a secret S is matched with the secret V.
1. A method for verifying the identity of a person based on a sequence of feature components extracted from a biometric sample from the person, comprising:
quantizing each feature component of said sequence of the feature components,
assigning a data bit sequence to each quantized feature component in such a way that data bit sequences for adjacent quantization intervals have a Hamming distance of 1, wherein a data bit sequence for one quantization interval exhibits a Hamming distance of 1 from a data bit sequence in a next quantization interval adjacent thereto,
concatenating said data bit sequence into a bit string with the data bit sequences assigned to each other quantized feature component,
combining said bit string with a helper data set using an exclusive disjunction (XOR) operation into a codeword,
decoding said codeword into a secret, and
matching a second secret with the decoded secret.
2. The method according to claim 1 , wherein each feature component is quantized into quantization intervals in such a way that the probability for each quantization interval is equal according to a statistical model.
3. The method according to claim 1 , wherein said decoding is made according to an error correcting code.
4. The method according to claim 3 , wherein said error correcting model is a binary error correcting code.
5. The method according to claim 4 , wherein said binary error correcting code is based on a Hamming code.
6. The method according to claim 3 , wherein said error correcting code is a multilevel error correcting code.
7. The method according to claim 6 , wherein said multilevel error correcting code is based on a BCH (Bose, Ray-Chaudhuri, Hocquenghem) code.
8. A method for creating a helper data set based on a sequence of feature components extracted from a biometric sample from a person and a secret to enable verifying the identity of the person, comprising:
quantizing each feature component of said sequence of the feature components,
assigning a data bit sequence to each quantized feature component in such a way that data bit sequences for adjacent quantization intervals have a Hamming distance of 1, wherein a data bit sequence for one quantization interval exhibits a Hamming distance of 1 from a data bit sequence in a next quantization interval adjacent thereto,
concatenating said data bit sequence into a bit string with the data bit sequences assigned to each other quantized feature component,
encoding said secret into a codeword, and
combining said bit string with the codeword into said helper data set by using an exclusive disjunction (XOR) operation.
9. A device for verifying the identity of a person based on a sequence of feature components extracted from a biometric sample from the person, comprising:
a quantizer for quantizing the feature components of said sequence of the feature components,
an assigner for assigning a bit sequence to each quantized feature component in such a way that data bit sequences for two adjacent quantization intervals have a Hamming distance of 1, wherein a data bit sequence for one quantization interval exhibits a Hamming distance of 1 from a data bit sequence in a next quantization interval adjacent thereto,
a concatenater for concatenating each bit sequence, output from said assigner, into a bit string,
an XOR-operator for combining said bit string with a helper data set into a codeword,
a decoder for decoding said codeword into a secret, and
a matcher for matching a second secret with the decoded secret.
10. A device for creating a helper data set based on a sequence of feature components extracted from a biometric sample from a person and a secret, comprising:
a quantizer for quantizing the feature components of said sequence of the feature components,
an assigner for assigning a data bit sequence to each quantized feature component in such a way that data bit sequences for two adjacent quantization intervals have a Hamming distance of 1, wherein a data bit sequence for one quantization interval exhibits a Hamming distance of 1 from a data bit sequence in a next quantization interval adjacent thereto,
a concatenater for concatenating each bit sequence, output from said assigner, into a bit string, and
an XOR-operator for combining said bit string with a corresponding codeword based on said secret into said helper data set.
11. A non-transitory computer readable medium having embodied thereon computer executable instructions that, when executed, implement a method for verifying the identity of a person based on a sequence of feature components extracted from a biometric sample from the person, the method comprising:
quantizing each feature component of said sequence of the feature components;
assigning a data bit sequence to each quantized feature component in such a way that data bit sequences for adjacent quantization intervals have a Hamming distance of 1, wherein a data bit sequence for one quantization interval exhibits a Hamming distance of 1 from a data bit sequence in a next quantization interval adjacent thereto;
concatenating said data bit sequence into a bit string with the data bit sequences assigned to each other quantized feature component;
combining said bit string with a helper data set using an exclusive disjunction operation into a codeword;
decoding said codeword into a secret; and
matching a second secret with the decoded secret.