Contactless hand biometrics trail 3D fingerprints in development for forensics use

Contactless hand biometrics are not accurate enough to be of much use on their own in forensic investigations at this time, but could have a place in victim or criminal investigation in combination with other approaches.

This conclusion is the product of a literature review and performance benchmark for the modality by researchers affiliated with the Biometrics and Security Research Group at Hochschule Darmstadt and the Technical University of Denmark.

Their paper breaks contactless hand biometrics down into hand part-based and whole-hand approaches. The relative merits of several algorithms are then compared.

Rank 1 identification rates reached as high as 96.8 percent, for easy and controlled scenarios with the MBA-Net approach. For the rest of the scenarios, the ABD-Net approach returned the best identification rate, either alone or tied, but cross-database IR topped out at 34.9 percent, and for the cross-database challenging scenario, it was only 11.2 percent.

The use of complimentary techniques, like soft biometrics, to reduce the reference dataset could yield useful results for forensics, the researchers say.

For authentication, contactless hand biometrics, the addition of modules such as for presentation attack detection and template protection are needed to elevate the state of the art to the point where it provides a practical security benefit.

Developing a cross-domain algorithm for 3D contactless fingerprints

Innovatrics, meanwhile, explains in a post to its website how it developed a technology that delivers a three-fold improvement in biometric accuracy with 3D fingerprints.

The company made gypsum casts of staff members’ fingerprint to understand how 3D scans of fingerprints distort the biometrics. The finger replicas were scanned with 3D imaging technology, and then the transformation of a finger photo into a flat fingerprint modeled.

Synthetic fingerprints were added to the dataset, and after months of research, Innovatrics’ data scientists developed a way to quickly and accurately render fingerprints, according to the post.

A rigorous series of tests followed, and the next step is for Innovatrics to build the final neural network into its mobile libraries and ABIS to enable its clients to benefit from contactless fingerprint biometrics.

NIST recently updated its guidance on contactless fingerprint interoperability.

Article Topics

biometric identification  |  biometrics  |  contactless biometrics  |  fingerprint recognition  |  forensics  |  Innovatrics