An evaluation of biometric fingerprint matchers in a forensic context using latent impressions

• Authors: Rainer Kärgel;Mario Hildebrandt;Jana Dittmann
• Venue: Proceedings of the on Multimedia and security
• Year: 2012

Exemplar fingerprints captured directly from a finger are used in biometric systems for authentication purposes. If the quality of the captured image is insufficient for extracting biometric features, it can be easily captured again. In forensics latent fingerprints from crime scenes have been analyzed for over a century. Here, the investigation process is usually performed manually by forensic experts. In addition to that, automated fingerprint identification systems (AFIS) can be used to identify potentially matching samples from a reference database. However, in forensics different kinds of errors might occur during the investigations. Our goal in this paper is to determine whether biometric fingerprint matchers can be used to aid forensic experts by investigating dusted latent fingerprints captured by a Canon EOS 550D digital camera with a resolution of 500 ppi. For that, a dataset of 380 latent and 100 exemplar prints is created and matched (LP-TP) using five freely available matching programs. Our overall goals are the evaluation of the classification speed and the achieved error rates for the matching of all fingerprints using different metrics, fingerprints of specific test persons, specific fingers as well as different substrates. Furthermore, we evaluate whether the NIST Fingerprint Image Quality (NFIQ) is suitable as an indicator for the matching accuracy. The results show that the matching speed is sufficient, but the error rates are too high for applying the matchers in a forensic context. The performance evaluation for the specific test persons, fingers and substrates indicates potential improvements by the fusion of different matchers. Based on our results, the NFIQ is not a suitable means of determining latent print quality.