Statistical Analysis of Heterogeneous Data in Forensic Science

• Activity Number: 287
• Type: Invited
• Date/Time: Tuesday, August 11, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Hari Iyer* and Soweon Yoon and Elham Tabassi
• Companies: National Institute of Standards and Technology and Michigan State University/National Institute of Standards and Technology and National Institute of Standards and Technology
• Keywords: Heterogeneous Data ; Likelihood Ratio ; Posterior probability ; Forensics ; Fingerprints ; Latent Prints

Abstract

Forensic science strives to provide objective and scientific approaches for collection and interpretation of crime scene data and as such forms the foundation for the criminal justice system in the U.S. With the exponential growth in information technology and data acquisition tools there is now a great need for improved and reliable statistical and probabilistic approaches for the treatment of forensic data.

One of the critical problems in forensic science is the evaluation of the evidential value of crime scene marks - the statistical measure of uncertainty in deciding whether or not a suspect being charged with a crime is in fact the source of the evidence. Such crime scene evidence may consist of one or more entities such as fingerprints, shoe marks, tire marks, blood stains, fiber samples, DNA samples, hair samples, etc. In this talk we mainly focus on the treatment of fingerprint data which consists of three levels of features: (i) coarse representation of pattern type and ridge flow (ii) spatial configuration of ridge endings and bifurcations, called minutiae, and (iii) fine details of friction ridges such as pores. We also briefly discuss shoeprints and tiremarks.