Digital forensics is an advancing field that finds and infers original evidence in the form of electronic data. More precisely, digital crime scene analysis complements this data, providing security while recording, documenting, and processing, and overall information management. Firearm identification in digital crime scene analysis is an integrated part of digital forensics dealing with firing pin shape matching. This paper introduces a multiple-slice-shape (MSS) approach to accurately process the firearm toolmark surface data.

The MSS approach comprises four parts: data acquisition, MSS preprocessing, MSS feature extraction, and classification. The first part uses “a confocal laser scanning microscope (CLSM) ... for contactless topography acquisition of firing pin impressions on central fire cartridge bottoms.” The second part consists of four separate steps: data preprocessing, slice segmentation, slice postprocessing, and corresponding point estimation. The third part employs multiple-angle-path (MAP) and multiple-circle-path (MCP) 3D spatial features to describe “change of local vertical gradient, or change of local lateral deviation in consideration of multiple angles,” and “straight or circular path lines applied to the complete firing pin impression or statistical analysis considering all slices,” respectively. The fourth part uses three classifiers, implemented through the WEKA data mining toolbox: a support vector classifier, a Bayesian net classifier, and a nearest neighbor classifier.

For experiments, the authors use a self-acquired test set containing three pairs of nine millimeter (mm) firearms (Walther P99, Ceska 75B, and Beretta 92FS), producing 72 cartridge samples. The MSS parameterization uses three angle counts {8,16,32} for MAP parameterization and three radius counts {10,15,20} for MCP parameterization. This approach achieved accuracy between 67 and 100 percent for the MAP feature and between 92 and 100 percent for the MCP feature.

In future work, the authors propose “an extended evaluation of parameter values” for improving classification accuracy, validating results for more individual guns and ammunition types, and addressing more global and local features.