Towards Quantification of the Weight of Evidence with Partial Fingermarks on Real Forensic Casework

In forensic fingerprint examinations, there is a need for statistical techniques to quantitatively assess the weight of the evidence and measure the performance of the fingerprint comparisons in more adequate ways that should be both empirical and scientific. The first step towards this evidence quantification for statistical reporting is to build evidence-evaluation methods that generate score from pairs of latent and impression images that the examiner has annotated with minutiae features manually. In real forensic casework scenario, the examiner manually compares the given latent fingerprint against a reference fingerprint and arrives at a logical conclusion of identification-exclusion decision based on examiner's training and experience following the ACE-V protocol, but recently this kind of procedures have been criticized, arguing for a more quantitative evaluation of the weight of the evidence to be produced in court. There is no scientific framework in use at the criminal justice system to characterize the uncertainty involved in the ACE-V procedure, as well as to express the strength of opinion of the forensic examiner quantitatively. Such a requirement has been articulated in several influential reports like the NRC 2009 report and the NIST Human Factors report. The new paradigm coming forward in this regard avoids hard identification decisions by considering evidence reporting methods that incorporate uncertainty and statistics. Among all the methods of evidence evaluation, the likelihood ratio is receiving greater attention. To use any statistics-based framework for quantification of evidence, scores are required at the ACE-V stage in place of logical decision (match, non-match or the comparison is inconclusive). We proposed a framework to generate a score from the matched latent template and the reference impression template at the ACE-V stage [1] [2]. Such a score can be utilized to quantitatively express the strength of opinion of the forensic examiner using statistics-based framework. Together with the description of the new realistic forensic casework driven score computation, we also exploited the developed framework to study the discriminating power of matched template [1] on the NIST Special Database (SD) 27 and the real forensic fingerprint database (GCDB) acquired from The Guardia Civil, the law enforcement agency of the Government of Spain. Along with the location and orientation attributes for minutiae, GCDB also consisted of type information for each minutiae. Apart from typical minutiae features (ridge-ending and bifurcations), GCDB also consisted of other rare minutiae features like fragments, enclosures, points/dots, interruptions etc. We also exploited the developed score computation framework to study the importance of rare minutiae features in the matched templates [2]. The results shows the feasibility of the developed approach towards quantification of the weight of evidence in forensic caseworks.