Indexação de impressões digitais utilizando uma abordagem textual

Abstract

The use of credentials such as passwords and codes can easily be lost and forgotten. For this reason, the use of biometric information in authenticating people has grown in recent years. Also, biometric information has a large number of application domains. Fingerprints are the most used information in applications that use biometric systems, being applied in mobile devices, banking systems, people identification systems in government agencies, and among others. With the increase in the number of people, more biometric information is stored in identification systems, making it necessary to use more robust indexing methods and searching for fingerprint data. Furthermore, textual information is generated faster, making it even more essential to use more robust parallel and distributed data processing techniques. For this reason, this work proposes the use of textual processing approaches applied to the problem of indexing and searching for fingerprints. Two techniques are implemented using textual data: Method 1 uses tables hash and locality sensitive hashing; Method 2 uses inverted index from Elasticsearch and locality sensitive hashing. Both methods obtained good results in the FVC 2002 DB1a data set, with 0.85% and 0.42% Error Rate on 10% Penetration Rate, respectively. However, in the experiment with the NIST SD14 dataset, the methods did not scale with the increase in fingerprints.