Performance Study of a Score-based Likelihood Ratio System for Forensic Fingermark Comparison.

Area of Science:

• Forensic Science
• Biometrics
• Pattern Evidence Analysis

Background:

• Likelihood ratio (LR) systems are crucial for evaluating evidence in forensic science.
• Automated Fingerprint Identification Systems (AFIS) are widely used for fingerprint comparisons.
• The evidential value of fingerprints with fewer than the traditional twelve points is a subject of ongoing research.

Purpose of the Study:

• To assess the performance of a score-based LR system for fingerprint comparisons.
• To evaluate the evidential strength of fingermarks with 6-11 minutiae.
• To determine the reliability of LR estimates and their uncertainty.

Main Methods:

• Utilized an AFIS comparison algorithm for score-based LR calculations.
• Evaluated hypotheses at the person level, not the finger level.
• Employed leave-one-out cross-validation and simulation studies to measure performance and interval coverage.
• Calculated bootstrap intervals to quantify sampling uncertainty.

Main Results:

• The score-based LR system demonstrated substantial evidential strength for same-source fingerprint comparisons.
• Significant evidential value was observed even for comparisons not meeting the Dutch twelve-point standard.
• Bootstrap intervals effectively indicated the sampling uncertainty of the LR estimates.

Conclusions:

• The studied LR system provides reliable and quantifiable measures of evidential strength in fingerprint comparisons.
• The findings support the use of LR systems for fingerprint evidence, including cases with limited minutiae.
• The methodology is adaptable for evaluating score-based LR systems in other forensic disciplines.