Biometric data is compared by algorithms, and therefore usually stored, in the form of templates. A biometric template is a mathematical representation of the features or characteristics of the source data, whether it is a fingerprint scan, facial image, or voice recording. In the registration process, source data is collected, the model is created, and then stored in a database as reference data.
Matching is performed by capturing probe data, converting it to a pattern in the same format as the reference pattern, and then comparing the two patterns for verification or identification.
Model building often involves pre-processing sample data, for example to improve contrast, before feature extraction. Sample data can be analyzed with quality assessment algorithms, such as NIST NFIQ 2 to determine its suitability for use in model building.
Hackers and thieves seek out biometric templates because if left unprotected, they could be used to impersonate the victim to gain access to valuable information or commit a crime. In addition to the risk of being stolen or hacked, potential vulnerabilities include replacement with an unauthorized pattern or tracking in various databases as a form of surveillance if associated with other personally identifiable information and improperly secured .
However, models are not images, and if properly encrypted, it is nearly impossible to reverse engineer images from images.
Model data can be encrypted once it enters a database, requiring a key to decrypt it and return it to a readable format.
In authentication or identity verification scenarios, biometric reference templates can be stored for one-to-one (1:1) matching in a discrete secure environment, such as the secure element of the mobile device. a user, although on-device pattern matching on its own does not provide assurance of a particular claimed identity.
Additional methods and technologies under development or relatively new to the market to protect models against theft or misuse include bio-hashing and model splitting.
Click here for more explanation of concepts in the field of biometrics.
biometric data | biometric matching | biometric template | biometrics | Data protection