Filing Information

Inventor(s)	William Gareth James HOWELLS · Michael Christopher FAIRHURST · Farzin DERAVI ·
Assignee(s)	University of Kent ·  View assignee updates
Attorney/Agent(s)	Nicholas Justin Ertl ·
Application Number	EP07804012.8
Filing date	07/03/2007
Issue date	10/27/2010
U.S. Classifications	--
International Classifications	G06K 9/62  ·
Kind Code	B1
PCT Application Number	GB2007002471 - 07/03/2007
PCT Publication Number	WO2008003945 - 01/10/2008

Abstract

Independent Claims | See all claims (27)

1. An encryption and decryption method, comprising: generating a code from a set of biometric samples, each sample in the form of a pattern, the code generating comprising: (i) analysing a plurality of samples and obtaining feature values for the samples, each sample being interpreted as a plurality of features of a previously defined feature set obtained during an initial calibration; (ii) normalising the feature values by adapting the feature values such that they may each take only a predetermined range of values, each with substantially equal likelihood; (iii) obtaining mean, variance and covariance values for the different feature values from the plurality of samples; (iv) setting quantization levels for each feature using the obtained mean and variance values, the quantization levels determining the range of feature values to be considered as equivalent during quantization; (v) de-correlating the feature values using de-correlating coefficients obtained from the covariance values and (vi) deriving a single code from a combination of the de-correlated feature values, wherein the single code comprises a sub-set of the bits derived from the combination of feature values; using the code to form a private encryption key; using the private encryption key to encrypt data. decrypting the encrypted data by taking a further biometric sample to generate the required private key, wherein the method is template-free having no physical record of the private key or any form of information regarding the biometric itself, wherein the initial calibration phase for determining the pre-defined feature set, comprises: obtaining multiple calibration samples form multiple users; analysing the calibration samples and obtaining feature values for the samples; obtaining mean, variance and covariance values for the different features from the plurality of samples, wherein those mean and variance values are used in setting static quantization intervals used to re-create each time dynamically said mean and variance values of said code generating phase.
2. A method as claimed in any preceding claim, wherein deriving a single value comprises: obtaining a de-correlated feature value for each feature; and combining the de-correlated feature values by: performing a summing operation to combine the de-correlated feature values into to a single word; removing the lowest significant bits of the word; and removing the highest significant bit of the word.
3. A computer program which wen run on a computer is adapted to perform the method as claimed in any preceding claim.
4. An encryption and decryption system, comprising: a system for generating a code from a set of biometric samples, comprising: means for analysing a plurality of samples and obtaining feature values for the samples, each sample being interpreted as a plurality of features of a previously defined feature set obtained during an initial calibration; processing means for : calculating mean and variance values for the different feature values from the plurality of samples; setting quantization levels for each feature using the obtained mean and variance values together with the sample data, the quantization levels determining the range of feature values to be considered as equivalent during quantization; normalising the feature values by adapting the feature values such that they may each take only a predetermined range of values, each with substantially equal likelihood; and deriving a single code from a combination of the feature values, wherein the single code comprises a sub-set of the bits derived from the combination of feature values; means for using the code to form a private encryption key; means for using the private encryption key to encrypt data. means for decrypting the encrypted data by taking a further biometric sample to generate the required private key, wherein the system is template-free having no physical record of the private key or any form of information regarding the biometric itself, wherein the system further comprises means for performing the initial calibration phase for determining the pre-defined feature set by: obtaining multiple calibration samples form multiple users; analysing the calibration samples and obtaining feature values for the samples; and obtaining mean, variance and covariance values for the different features from the plurality of samples, wherein those mean and variance values are used in setting static quantization intervals used to re-create each time dynamically said mean and variance values calculated by said system for generating a code.