Charles E.H. Berger

Evidence evaluation: A response to the court of appeal judgment in R v T

This is a discussion of a number of issues that arise from the recent judgment in R v T [1]. Although the judgment concerned with footwear evidence, more general remarks have implications for all disciplines within forensic science. Our concern is that the judgment will be interpreted as being in opposition to the principles of logical interpretation of evidence. We reiterate those principles and then discuss several extracts from the judgment that may be potentially harmful to the future of forensic science. A position statement with regard to evidence evaluation, signed by many forensic scientists, statisticians and lawyers, has appeared in this journal [2] and the present paper expands on the points made in that statement.

Signature Verification Competition for Online and Offline Skilled Forgeries (SigComp2011)

The Netherlands Forensic Institute and the Institute for Forensic Science in Shanghai are in search of a signature verification system that can be implemented in forensic casework and research to objectify results. We want to bridge the gap between recent technological developments and forensic casework. In collaboration with the German Research Center for Artificial Intelligence we have organized a signature verification competition on datasets with two scripts (Dutch and Chinese) in which we asked to compare questioned signatures against a set of reference signatures. We have received 12 systems from 5 institutes and performed experiments on online and offline Dutch and Chinese signatures. For evaluation, we applied methods used by Forensic Handwriting Examiners (FHEs) to assess the value of the evidence, i.e., we took the likelihood ratios more into account than in previous competitions. The data set was quite challenging and the results are very interesting.

Color separation in forensic image processing

ABSTRACT: In forensic image processing, it is often important to be able to separate a feature from an interfering background or foreground, or to demonstrate colors within an image to be different from each other. In this study, a color deconvolution algorithm that could accomplish this task is described, and it is applied to color separation problems in document and fingerprint examination. Subtle color differences (sometimes invisible to the naked eye) are found to be sufficient, which is demonstrated successfully for several cases where color differences were shown to exist, or where colors were removed from the foreground or background. The software is available for free in the form of an Adobe® Photoshop®‐compatible plug‐in.

The LR does not exist

More than 40years ago, De Finetti warned that probability is a misleading misconception when regarded as objectively existing exterior to the mind. According to De Finetti, probabilities are necessarily subjective, and quantify our belief in the truth of events in the real world. Given evidence of a shared feature of a trace and an accused, we apply this framework to assign an evidential value to this correspondence. Dividing 1 by the objectively existing proportion of the population sharing that feature would give that evidential value - expressed as a likelihood ratio (LR) - only if that proportion were known. As in practice the proportion can only be estimated, this leads some to project their sampling uncertainty - or precision - associated with the estimated proportion onto the likelihood ratio, and to report an interval. Limited data should limit our LR however, because as we will demonstrate the LR is given by what we know about the proportion rather than by the unknown proportion itself. Encapsulating all uncertainty - including sampling uncertainty of the proportion - our LR reflects how much information we have retrieved from the feature regarding the traces origin, based on our present knowledge. Not an interval but a number represents this amount of information, equal to the logarithm of the LR. As long as we know how to interpret the evidence with a well-defined probabilistic model, we know what our evidence is worth.

Objective paper structure comparison through processing of transmitted light images

A method for the comparison of paper structure using light transmission images and frequency analysis was developed. The resolution of the light transmission images and the algorithm for the feature extraction were greatly improved to enhance the visibility of peaks in the 2D power spectrum that results from frequency analysis. A comparison method based on correlation measures how well the spectra match as a function of the orientation of the paper, yielding an objective and quantitative measure of similarity between 0 and 1. A technical validation was carried out with 25 different papers showing the potential of this method with common copy papers. Finally, the method was applied in a case.

Measuring coherence of computer-assisted likelihood ratio methods

Measuring the performance of forensic evaluation methods that compute likelihood ratios (LRs) is relevant for both the development and the validation of such methods. A framework of performance characteristics categorized as primary and secondary is introduced in this study to help achieve such development and validation. Ground-truth labelled fingerprint data is used to assess the performance of an example likelihood ratio method in terms of those performance characteristics. Discrimination, calibration, and especially the coherence of this LR method are assessed as a function of the quantity and quality of the trace fingerprint specimen. Assessment of the coherence revealed a weakness of the comparison algorithm in the computer-assisted likelihood ratio method used.

Likelihood ratio methods for forensic comparison of evaporated gasoline residues

In the investigation of arson, evidence connecting a suspect to the fire scene may be obtained by comparing the composition of ignitable liquid residues found at the crime scene to ignitable liquids found in possession of the suspect. Interpreting the result of such a comparison is hampered by processes at the crime scene that result in evaporation, matrix interference, and microbial degradation of the ignitable liquid. Most commonly, gasoline is used as a fire accelerant in arson. In the current scientific literature on gasoline comparison, classification studies are reported for unevaporated and evaporated gasoline residues. In these studies the goal is to discriminate between samples of several sources of gasoline, based on a chemical analysis. While in classification studies the focus is on discrimination of gasolines, for forensic purposes a likelihood ratio approach is more relevant. In this work, a first step is made towards the ultimate goal of obtaining numerical values for the strength of evidence for the inference of identity of source in gasoline comparisons. Three likelihood ratio methods are presented for the comparison of evaporated gasoline residues (up to 75% weight loss under laboratory conditions). Two methods based on distance functions and one multivariate method were developed. The performance of the three methods is characterized by rates of misleading evidence, an analysis of the calibration and an information theoretical analysis. The three methods show strong improvement of discrimination as compared with a completely uninformative method. The two distance functions perform better than the multivariate method, in terms of discrimination and rates of misleading evidence.

Design and results of an exploratory double blind testing program in firearms examination

In 2010, the Netherlands Forensic Institute (NFI) and the University of Amsterdam (UvA) started a series of tests for the NFIs Firearms Section. Ten cartridge case and bullet comparison tests were submitted by various external parties as regular cases and mixed in the flow of real cases. The results of the tests were evaluated with the VU University Amsterdam (VUA). A total of twenty-nine conclusions were drawn in the ten tests. For nineteen conclusions the submitted cartridge cases or bullets were either fired from the questioned firearm or from one and the same firearm, in tests where no firearm was submitted. For ten conclusions the submitted cartridge cases or bullets were either fired from another firearm than the submitted one or from several firearms, in tests where no firearm was submitted. In none of the conclusions misleading evidence was reported, in the sense that all conclusions supported the true hypothesis. This article discusses the design considerations of the program, contains details of the tests, and describes the various ways the test results were and could be analyzed.

Implementing context information management in forensic casework: minimizing contextual bias in firearms examination

Managing context information in forensic casework aims to minimize task-irrelevant information while maximizing the task-relevant information that reaches the examiner. A design and implementation of context information management (CIM) is described for forensic firearms examination. Guided by a taxonomy of different sources of context information, a flow-chart was constructed that specifies the process of casework examination and context information management. Due to the risk of bias, another examiner may need to be involved when context information management is unsuccessful. Application of context information management does not make a subjective examination objective, but can limit the risks of bias with a minimal investment of time and resources.

Color Deconvolution and Support Vector Machines

Methods for machine learning (support vector machines) and image processing (color deconvolution) are combined in this paper for the purpose of separating colors in images of documents. After determining the background color, samples from the image that are representative of the colors to be separated are mapped to a feature space. Given the clusters of samples of either color the support vector machine (SVM) method is used to find an optimal separating line between the clusters in feature space. Deconvolution image processing parameters are determined from the separating line. A number of examples of applications in forensic casework are presented.