During latent print comparison, a non-mated print will be colloquially qualified as a close nonmatch (CNM) to the mark when it shows a level of agreement that could have misled an examiner to erroneously conclude that it was from the same source as the mark. These CNM prints are of crucial importance in developing the expertise of latent print examiners because they constitute the worst case scenario for a comparison between impressions originating from different sources. Because of the nature of the algorithms used in Automated Fingerprint Identification Systems (AFISs) (which are designed to find the closest matches contained in the database), combined with the ever-enlarging set of prints in their gallery, it is expected that examiners will increasingly face comparisons involving CNMs.
The issue facing the fingerprint community is that it does not have any repository of known cases including CNMs from which comparison exercises can be drawn. This is largely because CNMs are so time-consuming to locate. Over time, anecdotal cases have been shared in the community, but without any systematic organization to the data or means to search them available to law enforcement agencies. Additionally, these cases were submitted by examiners from casework, thus the ground truth state of the image pairs was unknown.
In this context, RTI International and University of Lausanne (Unil) present a proposal aimed at addressing CNM topics and delivering on the following objectives:
1. A ground truth library of marks and prints: Using an innovative partnership model to leverage an international network of public and private agencies and members of academia representing some of the top experts in the field, an international close nonmatch library (ICNML) of 1,000 cases will be prepared offering a repository of relevant and realistic marks and prints of known ground truth from 100 donors.
2. An evaluation of examiner understanding of CNMs: A list of red flags for CNMs will be developed based on expert input and discussion, and this list will be tested for its ability to predict the occurrence of CNMs. This information will increase our understanding of how well examiners understand CNMs and will allow for training recommendations to be made.
Developed using a crowd-sourcing approach to the research, the ICNML will be the first expert evaluated library of its type that will be made available to authorized agencies with a view toward promoting the training (and continuing education) of examiners, generating proficiency tests, providing images for use in research, and assisting in benchmarking AFIS systems.
This project contains a research and/or development component, as defined in applicable law, and complies with Part 200 Uniform Requirements - 2 CFR 200.210(a)(14).