Fingerprint evidence is a leading method for suspect identification

This Project Grant award of $111,000 from the National Institute of Justice (CFDA 16.560) to Iowa State University of Science and Technology will fund research to develop machine learning methods for extracting information from fingerprint samples. The first aim is to determine the time-since-deposition of a fingerprint, which could help investigators establish when a crime occurred. The second aim is to analyze fingerprint chemistry to infer the physical activity level of the individual, which could aid in identifying suspects. The research will leverage matrix-assisted laser desorption/ionization mass spectrometry and machine learning techniques to extract meaningful data from complex fingerprint samples that are currently too difficult to interpret. This work has the potential to provide valuable investigative insights from fingerprint evidence that is currently underutilized due to quality limitations.

Forensic investigators spend considerable time finding, enhancing, and collecting fingerprint evidence at crime scenes to match a person to the crime scene location. However, multiple issues arise when the fingerprint quality is too low to interpret (i.e., smeared or distorted) or if the fingerprint cannot be matched to a suspect in custody or using an online database, such as the fbis iafis. Both scenarios leave investigators with little information even though they have obtained trace evidence from a suspect. Consequently, over the last decade, research has focused on leveraging additional information from fingerprints that could aid investigations. These focuses can be narrowed to answering who committed the crime, when the crime was committed, and how the crime was committed. Proposed here are two research aims to look at method development with the intention to aid in answering when a crime occurred and who committed the crime using fingerprint chemical analysis. These methods will be developed on a matrix-assisted laser desorption/ionization mass spectrometer and will primarily focus on sebaceous fingerprints. To help determine when a crime occurred, a method to find a fingerprints time-since-deposition (TSD) will be the first aim. The second aim will focus on extracting information on the physical activity level of an individual based on their fingerprint. Knowing the physical activity level will allow law enforcement officers to focus their attention on certain occupations or habits the perpetrator has, thus aiding in finding who committed a crime. However, the chemical complexity of fingerprints has posed a significant obstacle in most fingerprint chemical analyses. Thus, machine learning will be used to ascertain information and patterns in the complex biological samples that are fingerprints. Many studies have moved to using machine learning as a valuable tool to analyze complex biological samples. Programs such as matlab make the development and testing of machine learning models straightforward and show great potential to be applicable to the forensic community. Hence, the proposed research will develop machine learning approaches to determine a fingerprints age and an individuals physical activity based on sebaceous fingerprints. Either would aid in providing valuable information to criminal justice officers from useable and previously unusable fingerprints. This research will result in accurate machine learning models for forensic scientists to determine a fingerprints TSD and the individuals physical activity. CA/NCF