Biometric cryptosystem

Recent research has investigated how to strongly combine biometrics and cryptography in order

• to enhance security and privacy with respect to the storage of biometric reference templates, and
• to employ biometrics as a wrapper mechanism for securely storing and revealing cryptographic keys.

However, this approach is challenged by slight variations within the measurement of biometrics. As subsequent submissions are subject to altered acquisition conditions, variations within position, physical execution and human interaction, their measurements are rarely identical.

This inherent fuzziness of biometric data gives rise to different cryptographic values of the same biometric trait. Thus, cryptographic techniques cannot be directly employed onto biometrics. However, by tolerating slight variability within its unlocking and locking sets, the fuzzy vault of Juels and Sudan [10] allows to intersect biometrics with a cryptographic concept, resulting in a biometric cryptosystem. This cryptographic concept allows an unordered set of values to be used for locking a secret key value inside the vault, such that someone who possesses a substantial number of locking elements will be able to unlock the secret. The secret value can be unlocked by sets that substantially overlap with the original locking set, which makes the fuzzy vault useful in circumstances where precise recall of the values that constitute the locking set is difficult or impossible. The inherent variance within measurements of biometric traits provides only approximate data and thus biometrics form a natural source of locking data for the fuzzy vault. Jain et al. [11; 12]] and several others have investigated the usage of biometric traits within the fuzzy vault. Compared with more conventional implementations of biometric authentication, the fuzzy vault provides the advantages of increased security and enhanced privacy by removing the need to store the biometric traits directly within the biometric system. Current biometric cryptosystems rely on Juels and Sudan’s fuzzy vault concept which restricts its employment to only one applicant or one biometric trait.

Considering that biometrics are public data and that confidential data is locked into the vault under their unique features, a set of multiple biometric inputs is expected to increase security in terms of spoofing biometrics. The complexity of an attack increases according to the number of biometrics involved. The purpose of our approach is to extend the single control fuzzy vault to a multiple control tool. This generalisation not only allows multiple users or multiple biometric traits to be granted access to a single secret, but also extends the vault’s capability to incorporate secret sharing scheme structures. Due to the similarities in employing polynomial interpolation, we identify a connection between Shamir’s secret sharing scheme [13] and the fuzzy vault. As pointed out by Juels and Sudan [10], it is not possible to allow fuzzy inputs in secret sharing by fuzzy hiding of individual shares in Shamir’s scheme. Nevertheless, we can borrow ideas of generalizing access structures in Shamir’s scheme in order to attain our goal.

In particular, the proposed construction within one of our publications of a "multiple control fuzzy vault" has been developed on a threshold, compartmented and hierarchical access structure. These constructions allow multiple biometrics to be bound with a secret given a secret sharing access structure. Systems based on our constructions offer the flexibility to choose among different biometric traits according to the environment. Those include scenarios applying multimodal biometric applications or shared biometric access control without requiring storage of unprotected biometric datasets. The multiple control fuzzy vault can thus be viewed to contribute significant new application opportunities.

More about Cryptography: "The fuzzy vault"

Publication: "The multiple-control fuzzy vault"