A biometric template is a set of stored biometric features. To create a template, we need to get a biometric sample through a capture device, e.g., a camera or fingerprint scanner. The captured biometric sample is further converted into a mathematical file. Thus, the template is a digital representation of biometric samples stored in a biometric database known as ABIS. The process of comparing templates from the database to the obtained samples is called biometric authorization or biometric identification.
The widespread deployment of biometric templates raises the question about the security of these applications. How can an impostor or fraudster potentially hack into a system or steal biometric data?
Since a template is a series of zeroes and ones in a mathematical file, a template is definitely harder to read and misuse than traditional passwords and tokens. However, there are several points vulnerable to hacking when it comes to templates:
1. The database where templates are stored can be hacked or stolen
2. The transmission of files from a capturing device to the server
3. Hosted environment with a third party server breach
A fingerprint template is the biometric data obtained by a fingerprint sensor. The process of obtaining a fingerprint template is called enrollment. During the enrollment, data is transformed into zeroes and ones, creating a mathematical file, which is further stored in the biometric database. These mathematical files are known as fingerprint templates although people are often confusing fingerprint templates with the images obtained during the capture of the biometric data.
In short, fingerprint templates are digital representations of captured fingerprints.
Biometric template formats are standardized into ANSI 381 for all saved fingerprint data and ISO 19794-2 format.
Biometric template matching is a machine vision technique that matches two templates and it is the very basic core of every biometric system. The matcher does not use original pictures. As a result, the template matcher produces a similarity score, which is a vector with a probability of two biometric templates matching each other. The lower the threshold, the higher the probability score of two templates matching.
Matching can be either used in the verification process – 1:1 comparison or identification process – 1:N search.
Typically, these are the typical problems of template matching:
– Background changes
– Background clutters