In today’s fast moving digital era, in which personal as well as business communication and complex transactions are taking on-the-go approach, traditional identification practices have proved to be unusable. Traditional identity practices like ID cards or documents were originally developed for human intervened identification and authentication, so they turn completely unusable for computers and IT systems. Enabling IT systems to read and verify an identity card will not be an efficient approach as it will carry inadequacies of current application to machines. Fortunately there is an identification method with proven track record of more than 100 years in forensics: personal identification by fingerprints. From shopping to communicating and banking to business transactions, identity verification has become a necessity to recognize people and save customers from fraud and other financial crimes. Identities can be easily protected with biometric identification and authentication.
Biometrics technology makes it possible to quickly identify or authentication an individual by leveraging processing power of new age computing and data storage and search capabilities of modern database systems. Streamlining user enrollment, identification of unknown identity and verification of a claimed identity are the main objectives behind a biometric system implementation.
User enrollment: establishing the user identity
User enrollment takes place prior to identification or authentication. User enrollment is the process of establishing user identity on a biometric system. It is like introducing new user on a biometric system so the system can identify the user in future should the need arise. An analogy to user enrollment would be like introducing a new employee to the security guard to allow her when she seeks access to the office building. User enrollment is often performed casually, however, it is an important process and success of the future user verification heavily depends on the quality of sample captured at the time of enrollment. A user has to present his biometric sample like fingerprint, iris, retina or palm vein to a biometric scanner or imaging device, which captures the biometric identifiers presented to it.
Originally captured sample data is in raw and not much usable form. To make it usable, it has to be post-process by pre-defined process in the system. This data is taken through processes like enhancement and compression to improve usability and resource efficiency. Eventually, a biometric template is generated out of the sample data, which is associated with demographic data of the person and stored in a biometric database, for future identification and authentication purposes.
In simple words: Seeking identity of an unknown sample, in a database of identities, is called Identification or an 1:N match. When a person’s biometric identifiers searched through the whole collection of biometric templates, it is called identification, as the identity is unknown and has to be search through the entire database. 1:N is a very common scenario in law enforcement and forensic applications, where identities are searched for the latent prints acquired from crime scenes. Identification process seeks the answer of the question “Who are you?” 1:N search may take from seconds to several hours depending on the size and efficiency of the database systems and computing ability of the IT systems. When database is large, a unique ID is often issued which can be used to perform 1:N match, i.e. the verification, which is explained in the following section.
Verification is the process of confirming a claimed identity. It answers the question: Are you who you claim you are? This process is called 1:1 match as the person already has his/her identity details (unlike identification process) and it needs to be verified by comparing with authenticate records. Unlocking your phone with registered fingerprints is a good example of identity verification or 1:1 match. The phone has to match the presented fingerprint just with the stored one, if it matches, identity gets verified. 1:1 match is also employed when number of records are large and might take considerable time using 1:N match. In 1:N method, a unique number or key (unique to that system) is also issued to the person, which is used to fetch the record first. Since the ID number can be stolen or shared, his biometrics is scanned to verify the authenticity of the claimed identity. The unique ID number quickly fetches the record of claimed identity from the database and the system does not have to search and compare with all the records. This method makes verification faster when biometric database has a lot of records.
India’s Aadhar biometric verification is a good example of 1:1 match. Aadhar is world’s largest biometric database of Indian citizens containing their iris and fingerprint templates. Citizens are provided with a unique number, which they present while seeking identity verification. The Aadhar number fetches the record in a matter of seconds and biometrics like iris or fingerprints are captured to verify the authenticity with that record. This process would have taken hours if Aadhar systems had employed 1:N match because Aadhar database consists of 1.25 billion records of biometric data of Indian nationals.
Segmented Identification (1:Few)
Segmented identification aka 1:Few match is a process in which a claimed or unclaimed identity is searched through a segment or records, on the basis of addition information presented/available. Additional information is used to fetch segment of records which match with the additional information. For example, a biometric database, which is searchable with demographic information as well, can be searched with date of birth. Date of birth will fetch a segment of records with same date of birth and then biometric sample is taken to compare against only the segmented records to perform verification. Additional information makes the search and compare process quick as the system only has to compare against a few records out of the whole database.
|Identification (1:N)||Verification (1:1)||Segmented Identification (1:Few)|
|Probe is searched and compared against all records||Probe is searched and compared only with record fetched with a unique identifier (e.g. ID Number)||Probe is searched and compared against a segment of records fetched with additional data (e.g. Date of Birth)|
|Can be slow if database size is large||Quick||Quicker than 1:N but slower than 1:1|
|Example: latent print search in forensics||Example: unlocking phone with stored fingerprint||Searching a patient record with DoB and verifying with her fingerprint|
Despite the increasing hardware and software efficiency along with processing power of information systems, numbers can always outshine computing ability. When number of records grows exponentially, processing power may fall short and more powerful systems are required. However, search techniques can always expedite the process. Aadhar biometric database consists of 1.25 billion iris, fingerprints and photographs of Indian citizens, searching through this gigantic database could have not only been inefficient, it would have also been heavy on resources. Using unique identification number solved this problem and record of claimed identity can be fetched in the matter of seconds. As the world becomes a global villages, and number of biometric records increases, we may require more efficient search methods as well as computing ability.