Search

Since the 1960s, there have been calls for forensic voice comparison to be empirically validated under casework conditions. Since around 2000, there have been an increasing number of researchers and practitioners who conduct forensic-voice-comparison research and casework within the likelihood-ratio framework. In recent years, this community of researchers and practitioners has made substantial progress toward validation under casework conditions becoming a standard part of practice: Procedures for conducting validation have been developed, along with graphics and metrics for representing the results, and an increasing number of papers are being published that include empirical validation of forensic-voice-comparison systems under conditions reflecting casework conditions. An outstanding question, however, is: In the context of a case, given the results of an empirical validation of a forensic-voice-comparison system, how can one decide whether the system is good enough for its output to be used in court? This paper provides a statement of consensus developed in response to this question. Contributors included individuals who had knowledge and experience of validating forensic-voice-comparison systems in research and/or casework contexts, and individuals who had actually presented validation results to courts. They also included individuals who could bring a legal perspective on these matters, and individuals with knowledge and experience of validation in forensic science more broadly. We provide recommendations on what practitioners should do when conducting evaluations and validations, and what they should present to the court. Although our focus is explicitly on forensic voice comparison, we hope that this contribution will be of interest to an audience concerned with validation in forensic science more broadly. Although not written specifically for a legal audience, we hope that this contribution will still be of interest to lawyers.

For several decades, legal and scientific scholars have argued that conclusions from forensic examinations should be supported by statistical data and reported within a probabilistic framework. Multiple models have been proposed to quantify and express the probative value of forensic evidence. Unfortunately, the use of statistics to perform inferences in forensic science adds a layer of complexity that most forensic scientists, court officers and lay individuals are not armed to handle. Many applications of statistics to forensic science rely on ad-hoc strategies and are not scientifically sound. The opacity of the technical jargon used to describe probabilistic models and their results, and the complexity of the techniques involved make it very difficult for the untrained user to separate the wheat from the chaff. This series of papers is intended to help forensic scientists and lawyers recognize limitations and issues in tools proposed to interpret the results of forensic examinations. This article focuses on tools that have been proposed to leverage the use of similarity scores to assess the probative value of forensic findings. We call this family of tools ‘score-based likelihood ratios’. In this article, we present the fundamental concepts on which these tools are built, we describe some specific members of this family of tools, and we compare them explore to the Bayes factor through an intuitive geometrical approach and through simulations. Finally, we discuss their validation and their potential usefulness as a decision-making tool in forensic science.

For several decades, legal and scientific scholars have argued that conclusions from forensic examinations should be supported by statistical data and reported within a probabilistic framework. Multiple models have been proposed to quantify and express the probative value of forensic evidence. Unfortunately, the use of statistics to perform inferences in forensic science adds a layer of complexity that most forensic scientists, court offices and lay individuals are not armed to handle. Many applications of statistics to forensic science rely on ad hoc strategies and are not scientifically sound. The opacity of the technical jargon that is used to describe these probabilistic models and their results, and the complexity of the techniques involved make it very difficult for the untrained user to separate the wheat from the chaff. This series of article is intended to help forensic scientists and lawyers recognize limitations and issues in tools proposed to interpret the results of forensic examinations. This article focuses on the tool proposed by the Latent Print Branch of the U.S. Defense Forensic Science Center (DFSC) and called FRStat. In this article, I explore the compatibility of the results outputted by FRStat with the language used by the DFCS to report the conclusions of their fingerprint examinations, as well as the appropriateness of the statistical modelling underpinning the tool and the validation of its performance.

Prior work has shown that the contact surfaces of footwear rapidly lose very small particles (VSP) when walking on dry soil. Other research, with more general sampling from shoe soles, has shown that particles can persist much longer. Given rapid losses from contact surfaces, we hypothesize that non-contact, recessed areas of footwear retain particles from prior exposures. Here we conduct differential analysis of VSP recovered separately from contact and recessed surfaces, following a controlled series of environmental exposures. Work boots and tennis shoes were exposed by walking distances of 250 m, sequentially, in three environmental sites. VSP were harvested separately from contact and recessed surfaces using a moist swabbing procedure. VSP were analyzed by microscopy and the proportions attributable to each site were determined using a newly developed statistical model. The principal findings are (1) contact surfaces of footwear are dominated by VSP attributable to the most recent site of exposure, (2) recessed surfaces of footwear retain VSP from prior exposures, (3) sole type appears to be a source of major differences in the amount of VSP from prior exposures remaining in recessed areas, and (4) when VSP attributable to prior exposures are found, there is no clear trend for dominance between earlier exposures. The significance of these findings is considerable. In cases where the last site visited is of interest, VSP from contact surfaces of footwear will give a nearly pure sample of that site. In cases where prior sites visited are of interest, the VSP from the contact surfaces can be used as a background signal, providing a means for differential analysis of VSP mixtures found in the recessed areas of the sole. The strong influence of shoe sole type on the retention of VSP from prior exposures indicates that efforts to explore the rates of VSP loss from recessed areas will need to take characteristics of the sole type into account. The absence of a clear trend for dominance between earlier exposures suggests that once traces are lodged within the recessed areas of footwear soles they can be retained for long periods of time. These findings support prior research on the retention of particles on footwear that shows periods of long retention following rapid initial losses of trace materials. The differential analysis of VSP is a significant contribution to the methods available for the deconvolution of particle mixtures into fractions that are characteristic of their contributing sources.

In this correction, we provide an acknowledgment for the source of several images in the paper: J. Hendricks, C. Neumann, C.P. Saunders (2021) “Quantification of the weight of fingerprint evidence using a ROC-based Approximate Bayesian Computation algorithm for model selection”, Electronic Journal of Statistics (15)1, pp. 1228–1262.

For more than a century, fingerprints have been used with considerable success to identify criminals or verify the identity of individuals. The categorical conclusion scheme used by fingerprint examiners, and more generally the inference process followed by forensic scientists, have been heavily criticised in the scientific and legal literature. Instead, scholars have proposed to characterise the weight of forensic evidence using the Bayes factor as the key element of the inference process. In forensic science, quantifying the magnitude of support is equally as important as determining which model is supported. Unfortunately, the complexity of fingerprint patterns render likelihood-based inference impossible. In this paper, we use an Approximate Bayesian Computation (ABC) model selection algorithm to quantify the weight of fingerprint evidence. We supplement the original ABC model selection algorithm using a Receiver Operating Characteristic curve to mitigate its known shortcomings with respect to the choice of a suitable threshold and the curse of dimensionality. In a sense, we offer an alternative to other methods that have tried to address the same issues. Our method is straightforward to implement, computationally efficient, and visually intuitive for lay individuals (i.e., jurors). We apply our method to quantify the weight of fingerprint evidence in forensic science, but we note that it can be applied to any other forensic pattern evidence.

This Letter to the Editor is a reply to Mohammed et al. (2021) https://doi.org/10.1016/j.fsisyn.2021.100145 , which in turn is a response to Morrison et al. (2020) “Vacuous standards – subversion of the OSAC standards-development process” https://doi.org/10.1016/j.fsisyn.2020.06.005 .

Fingerprint evidence has been used to support the identification of criminals for more than a century. Statistical analysis of information present on finger impressions is not a new endeavour. Sir Francis Galton is credited with the first extensive analysis of fingerprint variability. The verification process relies primarily on good quality impressions. This allows for most of the requests for verification to be processed entirely automatically, with a response time of a few minutes, if not seconds. Latent prints that have been deemed suitable for further examination are then compared to control impressions from known donors. Subjectivity plays a critical role in the conclusions resulting from fingerprint examinations. Different countries have different rules regarding the quality and quantity of features that need to correspond between the trace and control impressions. Bayesian reasoning may require a significant paradigm shift in what examiners would be allowed to report and how they would report it.

This project used established analytical tools and statistical methods to determine the evidential value of very small particle (VSP) profiles found on handguns, cell phones, drug packaging, and ski masks. Sampling protocols were designed, tested and used to sample VSP from evidence items from a single county-level crime laboratory: 30 handguns, 31 cell phones, 36 drug packaging specimens and 32 ski masks. Specimens were prepared for analysis employing established protocols for semi-automated scanning electron microscopy with elemental characterization by energy dispersive x-ray analysis (SEM/EDS). Statistical methods of particle combination analysis were applied to (1) remove particle "noise" from the datasets, (2) define a set of highly discriminating target particle types, (3) measure the strength of correspondence between profiles, and (4) measure the potential of VSP as an evidence type under defined experimental conditions. Most (84%) of the VSP specimens recovered from common evidence items showed sufficient variety and complexity in their VSP profiles to allow meaningful classification among closed sets of approximately 30 specimens. Correct associations were achieved for 93.5% of test specimens (drug packaging: 97.2%; cell phones: 92.6%; handguns: 92.9%; ski masks 88.2%). Test specimens with VSP numbers greater than 125 showed predominantly correct classifications. These findings establish (1) that VSP are present on the surfaces common items of physical evidence, (2) that the VSP can be efficiently recovered, prepared and analyzed by computer-assisted SEM/EDS analysis, (3) that the variety of particles is sufficient for the definition of classifiers based on reference sources, and (4) that the classifiers perform very well for these particle sets, showing that VSP recovery, analytical methods and computational methods are working effectively. The use of adhering VSP to establish quantitative associations among items of physical evidence is a new approach, exploiting a form of trace evidence that is typically ignored. It is highly significant for its potential to expand the number of cases to which trace evidence can meaningfully contribute and for its ability to include a quantitative statistical approach to data interpretation.

Recent developments in forensic science have lead to a proliferation of methods for quantifying the probative value of evidence by constructing a Bayes Factor that allows a decision-maker to select...

The objective comparison of complex signals in chemistry, and more particularly in forensic chemistry, with the view of inferring the source of a particular ‘trace’ object is an ongoing issue. In this paper, we propose a method that enables to assign a probability distribution to control material from any given source based on its chemical signal and to subsequently infer the source of a trace object using a simple Bayes classifier. Our method takes advantage of the dimension reduction and discriminative powers of kernels, and only requires the estimation of three parameters (once a kernel is chosen). We illustrate the application of our method to the inference of the source of trace objects based on very small particles (VSP) that can be found on their surfaces. VSPs are picked up in the environment(s) where the trace object has been in. These VSPs can (1) offer information about the geographic origins of the objects; (2) help discriminate between multiple mass-manufactured objects that would be otherwise identical. In this project, we use VSPs recovered from carpet fibers throughout the United States and apply our method to (1) reduce the complexity of compositional data obtained by SEM/EDS; (2) infer the source of the trace material. This method can be extended to VSPs found on other types of recovered forensic materials such as weapons, drugs, or IEDs (improvised explosive device), and to other types of chemical signals.

While most evidence types considered by forensic scientists result from the interactions between criminals, objects or victims at crime scenes, dust evidence arises from the mere presence of individuals and objects at locations of interest. Dust is ubiquitous. Yet, the use of dust evidence is anecdotical and is limited to cases where rare and characteristic particles are observed. The dust at any given location contains a large number of particles from different types and the dust present on an object or individual traveling across locations may be indicative of the locations recently visited by an individual, and, in particular, of the presence of an individual at a particular site of interest, e.g., the scene of a crime. In this paper, we propose to represent dust mixtures as vectors of counts of the individual particles, which can be characterised by any appropriate analytical technique. This strategy enables us to describe a dust mixture as a mixture of multinomial distributions over a fixed number of dust particle types. Using a latent Dirichlet allocation model, we make inference on (a) the contributions of sites of interest to a dust mixture, and (b) the particle profiles associated with these sites.

Over the past decade, the field of forensic science has received recommendations from the National Research Council of the U.S. National Academy of Sciences, the U.S. National Institute of Standards and Technology, and the U.S. President's Council of Advisors on Science and Technology to study the validity and reliability of forensic analyses. More specifically, these committees recommend estimation of the rates of occurrence of erroneous conclusions drawn from forensic analyses. “Black box” studies for the various subjective feature-based comparison methods are intended for this purpose. In general, “black box” studies often have unbalanced designs, comparisons that are not independent, and missing data. These aspects pose difficulty in the analysis of the results and are often ignored. Instead, interpretation of the data relies on methods that assume independence between observations and a balanced experiment. Furthermore, all of these projects are interpreted within the frequentist framework and result in point estimates associated with confidence intervals that are confusing to communicate and understand. We propose to use an existing likelihood-free Bayesian inference method, called Approximate Bayesian Computation (ABC), that is capable of handling unbalanced designs, dependencies among the observations, and missing data. ABC allows for studying the parameters of interest without recourse to incoherent and misleading measures of uncertainty such as confidence intervals. By taking into account information from all decision categories for a given examiner and information from the population of examiners, our method also allows for quantifying the risk of error for the given examiner, even when no error has been recorded for that examiner. This opens the door to the detection of behavioural patterns in the decision-making of examiners through their ABC rate estimates. These patterns could be used to detect error prone examiners, enabling additional training efforts to be more tailored to each examiner, limiting the risk of errors before they occur. We illustrate our proposed method by reanalysing the results of the “Noblis Black Box” study by Ulery et al. [ 18 ]. We did not choose this study because we disagree with their results, but because it is a good example of a study with dependent observations and missing data, and the data is publicly available. The ABC estimates for the population generally agreed with Ulery et al.’s plug-in estimates. However, credible intervals obtained from ABC are much wider than the confidence intervals for the corresponding parameter estimates that did not account for the dependencies among observations.

In the various forensic science disciplines, recent analytical developments paired with modern statistical computational tools have led to the proliferation of adhoc techniques for quantifying the probative value of forensic evidence. Many legal and scientific scholars agree that the value of evidence should be reported as a likelihood ratio or a Bayes Factor. Quantifying the probative value of forensic evidence is subjected to many sources of variability and uncertainty. There is currently a debate on how to characterize the reliability of the value of evidence. Some authors have proposed associating a confidence/credible interval with the value of evidence assigned to a collection of forensic evidence. In this paper, we will discuss the reasons for our opinion that interval quantifications for the value of evidence should not be used directly in the Bayesian decision-making process to determine the support of the evidence for one of the two competing hypotheses.

Common carp is an invasive species in North American waters, disrupting ecological systems and replacing native species in water systems they enter. Mass removal by seining of this species from infected waters is a labor-intensive task that requires knowledge of their behavior and popular aggregation locations. It has been shown that carp tend to aggregate and slow down during winter months leaving them potentially vulnerable to seining. In 2010, Hennen used a fixed telemetry system to track carp movement during ice-cover periods and reported on their spatial distributions. Expanding on this work, we propose a model to describe the discrete movement of carp through a fixed telemetry system using a Bayesian Hierarchical Markov model.

Fingerprints have been a valuable method of personal identification for more than 100 years due to the considerable variability in fingerprint patterns among individuals. Fingerprints were initiall...

An abundant scientific and legal literature, as well as several reports from the U.S. National Academy of Sciences, demands that forensic findings be supported by data and that conclusions be logic...

The aim of this review is to summarize the forensic science and analytical chemistry literature related to the comparison of infrared spectra. Most of the peer-reviewed papers referenced in this review have been published during the past ten years. This review is divided into two major (although not all-inclusive) sections: library searching (comparing one spectrum or group of spectra representing a sample to a spectral database) and what has been called “pattern recognition” (differentiating one set of spectra from another set by identifying the spectral features that provide the necessary discrimination). In the absence of discriminating features, spectra may be determined to not be “significantly different” and, in the forensic context, the possibility of the spectra originating from the same source (e.g., assembly plant in the case of an automotive paint sample) cannot be discarded. Library searching algorithms report a “quality index” or “similarity score,” whereas pattern recognition techniques attempt to collect information from a training set in the form of a response function that can be applied independently to other spectra for the purpose of classification, discrimination or even association. Additional tools are also discussed within this review including the use of a likelihood ratio to estimate the strength of similarity between one spectrum and another. In forensic applications, library search algorithms and pattern recognition techniques can only provide class evidence and cannot individualize the sample, e.g., in the case of an automotive paint search, they cannot identify the specific vehicle, only the “make” and model of the vehicle.

It has previously been shown that a walk of 250 m (approximately 175 steps/shoe) in a new environment is sufficient to remove and replace particles present on the contact surfaces of footwear. However, it is unknown how quickly this replacement occurs. This paper describes experiments measuring how quickly (in terms of steps) this loss and replacement of VSP occurs. Three environmental exposure sites were chosen to have different, characteristic particle types (soil minerals). Footwear of two types (work boots and tennis shoes) were tested, accumulating particles by walking 250 m in one site, followed by walking a designated number of steps (per shoe) in a second site. Very small particles (VSP) were harvested from contact surfaces of the footwear (those surfaces in direct contact with the ground when walking) using a moist swabbing procedure. The resulting numbers and types of VSP were determined using forensic microscopy and the proportions of VSP attributable to the first and second site were determined by particle combination analysis using a latent Dirichlet allocation model. The principal findings from this study are (1) that the contact surfaces of footwear are dominated by VSP attributable to the most recent site of exposure, (2) that walking in a new location rapidly removes and replaces VSP from the contact surfaces of footwear, (3) major replacement occurs in 5 to 10 steps and nearly complete replacement occurs by 25 steps, (4) the character of the loading site may influence the initial rate of loss and replacement of VSP (during the first five steps), and (5) overall, the loss and replacement of VSP on footwear in these experiments can be reasonably described by the relationship 1/(Steps + 2.68). The significance of these findings is considerable. In cases where the last site visited is of interest, VSP from contact surfaces of footwear will give a nearly pure sample of that site. In cases where prior sites visited are of interest, the VSP from the contact surfaces can be used as a background signal, providing a necessary beginning for efforts to resolve mixtures of VSP arising from different exposures. The usefulness of VSP on footwear evidence depends on developing this ability so that a reliable, relevant evidentiary “signal” can be separated from background noise (or signals from other exposures).

Environmentally acquired very small particles (VSP), present on the surfaces of carpet fibers, have shown potential for the association of fibers with their carpet source. To unlock this potential, research is required addressing a number of areas, including the application of methods under realistic casework conditions and the utilization of computational methods for the refinement and testing of the approach. In this work field collections of carpet fibers were conducted by crime scene practitioners under realistic casework conditions. VSP were isolated using previously developed methods, and analyses were conducted using SEM/EDS analytical protocols in an operational crime laboratory setting. Computational methods were designed, allowing sets of hundreds to thousands of VSP to be characterized. Classifiers were designed to associate and discriminate among specimens. These classifiers were applied to the VSP data for specimens collected by crime scene practitioners, as well as to a previously collected research dataset. Quantitative measures of correspondence and probative value were designed based on the classification measures and successfully applied to both sets of VSP data. Particle sets larger than 500 showed strong promise for quantitative associations with their sources. The use of larger numbers of target particle types (TPTs) showed strong promise to improve the performance of classification and association. Overall, the usefulness of VSP to provide objective, quantitative associations has been established. Because VSP are acquired post-manufacture, these methods can address fundamental limitations to probative value that arise when class characteristics, determined by manufacture, are shared among mass produced commodities. These findings are of broad significance for the future of trace evidence analysis. The results of this research are likely extendable, with minor modifications, to other trace evidence types (such as glass, tape and human hair), and are expected to contribute significantly for those types of trace evidence that have long been considered of low evidential value (such as undyed cotton and animal hairs). Furthermore, entirely new approaches to trace evidence are enabled by exploiting VSP profiles, such as comparing different types of trace evidence with one another and comparing VSP defined by crime scene or suspect environments to those on virtually any item of physical evidence.

This research focuses on detection of bi-allelic insertion/deletion polymorphisms (InDels) from challenged samples using the Investigator DIPplex® Kit from Qiagen. The study included analyzing body fluids from humans, as well as pristine and degraded samples. For the purpose of assessing species specificity, samples from various animals were included. At first, an analytical threshold (AT) for the detection of alleles was established based on an assessment of the noise in the system. Then, InDel profiles were obtained from samples exposed to detrimental environmental conditions, washed bloodstains, lipsticks, ChapStick®, ancient Croatian bone samples, and every day products such as toothbrushes and dental floss. Concordant profiles were obtained from different body fluids of the same donor. InDel profiles were also generated successfully when body fluids were deposited on substrates and directly amplified without pre-treatment with buffer or washing reagents. InDels can provide additional information when only partial STR profiles are generated from challenged samples.

Summary The fingerprint has, with considerable justification, come to be regarded as the acme of forensic identification. Over the last century, millions of cases have been resolved world wide because of marks left at crime scenes. The comparison methodology has not evolved greatly during its history and it is universal practice to present fingerprint evidence to a court as a categoric opinion of identification or exclusion, or to classify the evidence as inconclusive and not to report it. There has been a growing movement to supplement the fingerprint examination process by one that has a statistical model, supported by appropriate databases for calculating numerical measures of weight of evidence. The movement calls for the establishment of a logical framework for informing conclusions, based on explicit assumptions and data and open to revision and improvement. The aim is to enable the numerical evaluation of evidence that would currently be reported as a categorical identification and also of evidence that would currently be classified as inconclusive. The paper presents the results of a project carried out by the Forensic Science Service that aims to attain this goal. After a historical review, we describe a formal model for assigning numerical values to configurations of minutiae in fingerprints. We describe how the parameters of the model have been optimized to take account of interoperator variability and distortion of the finger pad, and we present the results of a substantial validation experiment that was based on searches that have been carried out on the US national fingerprint database of approximately 600 million fingerprints.

Fingerprints have been used for a century to identify criminals. But, astonishingly, fingerprint experts rely on subjective opinion, not on objective science. Yet they are required to claim absolute certainty for their judgements – a certainty that is mythical. Cedric Neumann brings probabilities and the hope of better justice to the courtroom; with Julian Champkin he explains the idea.

Research projects aimed at proposing fingerprint statistical models based on the likelihood ratio framework have shown that low quality finger impressions left on crime scenes may have significant evidential value. These impressions are currently either not recovered, considered to be of no value when first analyzed by fingerprint examiners, or lead to inconclusive results when compared to control prints. There are growing concerns within the fingerprint community that recovering and examining these low quality impressions will result in a significant increase of the workload of fingerprint units and ultimately of the number of backlogged cases. This study was designed to measure the number of impressions currently not recovered or not considered for examination, and to assess the usefulness of these impressions in terms of the number of additional detections that would result from their examination.

Shoemark evidence remains a cornerstone of forensic crime investigation. Shoemarks can be used at a crime scene to reconstruct the course of events; they can be used as forensic intelligence tool to establish links between crime scenes; and when control material is available, used to help infer the participation of given individuals to the commission of a crime. Nevertheless, as for most other impression evidence, the current process used to evaluate and report the weight of shoemark evidence is under extreme scrutiny. Building on previous research, this paper proposes a model to evaluate shoemark evidence in a more transparent manner. The model is currently limited to sole pattern and wear characteristics. It does not account formally for cuts and other accidental damages. Furthermore, it requires the acquisition of relevant shoemark datasets and the development of automated comparison algorithms to deploy its full benefits. These are not currently available. Instead, we demonstrate, using casework examples, that a pragmatic consideration of the various variables of the model allows us to already evaluate shoemark evidence in a more transparent way and therefore begin to address the current scientific and legal concerns.

The authors have published elsewhere a quantitative method for assessing weight of evidence in the case where a finger mark from a crime scene is compared with a control print taken from a single finger of a suspect. The approach is based on the notion of calculating a likelihood ratio (LR) that addresses a pair of propositions relating to the single finger that was the origin of the crime mark. In practice, things are rather different because the crime mark will not just be compared with a single finger from a suspect but with a set of prints from all of his/her fingers; likewise, when the mark is compared with a database, this will consist of ten print records from random individuals. It is clear that “finger propositions” are not realistic in this situation and we show how our approach may be generalised to address a pair of propositions that relate to the person that made the crime mark. It often is the case that information is present at the crime scene that enables some inference to be drawn relating to which of the offender's ten fingers left a particular mark of interest. This kind of inference may profitably be drawn into the formal analysis. We illustrate our approach with an example.

The ASTM standards on Writing Ink Identification (ASTM 1789-04) and on Writing Ink Comparison (ASTM 1422-05) are the most up-to-date guidelines that have been published on the forensic analysis of ink. The aim of these documents is to cover most aspects of the forensic analysis of ink evidence, from the analysis of ink samples, the comparison of the analytical profile of these samples (with the aim to differentiate them or not), through to the interpretation of the result of the examination of these samples in a forensic context. Significant evolutions in the technology available to forensic scientists, in the quality assurance requirements brought onto them, and in the understanding of frameworks to interpret forensic evidence have been made in recent years. This article reviews the two standards in the light of these evolutions and proposes some practical improvements in terms of the standardization of the analyses, the comparison of ink samples, and the interpretation of ink examination. Some of these suggestions have already been included in a DHS funded project aimed at creating a digital ink library for the United States Secret Service.

The research reported in this series of article aimed at (1) automating the search of questioned ink specimens in ink reference collections and (2) at evaluating the strength of ink evidence in a transparent and balanced manner. These aims require that ink samples are analysed in an accurate and reproducible way and that they are compared in an objective and automated way. This latter requirement is due to the large number of comparisons that are necessary in both scenarios. A research programme was designed to (a) develop a standard methodology for analysing ink samples in a reproducible way, (b) comparing automatically and objectively ink samples and (c) evaluate the proposed methodology in forensic contexts. This report focuses on the last of the three stages of the research programme. The calibration and acquisition process and the mathematical comparison algorithms were described in previous papers [C. Neumann, P. Margot, New perspectives in the use of ink evidence in forensic science-Part I: Development of a quality assurance process for forensic ink analysis by HPTLC, Forensic Sci. Int. 185 (2009) 29-37; C. Neumann, P. Margot, New perspectives in the use of ink evidence in forensic science-Part II: Development and testing of mathematical algorithms for the automatic comparison of ink samples analysed by HPTLC, Forensic Sci. Int. 185 (2009) 38-50]. In this paper, the benefits and challenges of the proposed concepts are tested in two forensic contexts: (1) ink identification and (2) ink evidential value assessment. The results show that different algorithms are better suited for different tasks. This research shows that it is possible to build digital ink libraries using the most commonly used ink analytical technique, i.e. high-performance thin layer chromatography, despite its reputation of lacking reproducibility. More importantly, it is possible to assign evidential value to ink evidence in a transparent way using a probabilistic model. It is therefore possible to move away from the traditional subjective approach, which is entirely based on experts' opinion, and which is usually not very informative. While there is room for the improvement, this report demonstrates the significant gains obtained over the traditional subjective approach for the search of ink specimens in ink databases, and the interpretation of their evidential value.

Fingerprints have been used with considerable success over the past century in multiple civil government, law enforcement and criminal investigation applications. Since the 1970s, computer assisted systems (AFIS - Automated Fingerprint Identification System) have been increasingly used to automatically compare fingerprints and propose associations between multiple friction ridge impressions of known or unknown sources. AFIS were initially entirely subordinated to human examiners. Improvements in the matching algorithms and workload considerations have pushed agencies to implement completely automated processes, known as "Lights-out" modes, where AFIS's render unsupervised conclusions on the donors of the queries. Such fully automated process is common for tenprint-to-tenprint comparisons; however it is currently not being widely adopted for other types of comparisons, such as latent print-to-tenprint comparisons. In this paper, we explore a statistical model that can be used to facilitate the latent print examination workflow by predicting whether any latent print should be searched in AFIS. In particular, we are interested in preventing poor quality latent prints to be searched in vain, and thus unnecessarily consume resources. Ultimately, we show that our model could be used to efficiently manage workflow and workload by categorizing latent prints as a function of the quality and quantity of information that can be observed on them, which enables examiners to select the most appropriate examination and quality assurance processes for each print.

Recent court challenges have highlighted the need for statistical research on fingerprint identification. This paper proposes a model for computing likelihood ratios (LRs) to assess the evidential value of comparisons with any number of minutiae. The model considers minutiae type, direction and relative spatial relationships. It expands on previous work on three minutiae by adopting a spatial modeling using radial triangulation and a probabilistic distortion model for assessing the numerator of the LR. The model has been tested on a sample of 686 ulnar loops and 204 arches. Features vectors used for statistical analysis have been obtained following a preprocessing step based on Gabor filtering and image processing to extract minutiae data. The metric used to assess similarity between two feature vectors is based on an Euclidean distance measure. Tippett plots and rates of misleading evidence have been used as performance indicators of the model. The model has shown encouraging behavior with low rates of misleading evidence and a LR power of the model increasing significantly with the number of minutiae. The LRs that it provides are highly indicative of identity of source on a significant proportion of cases, even when considering configurations with few minutiae. In contrast with previous research, the model, in addition to minutia type and direction, incorporates spatial relationships of minutiae without introducing probabilistic independence assumptions. The model also accounts for finger distortion.

Recent challenges to fingerprint evidence have brought forward the need for peer-reviewed scientific publications to support the evidential value assessment of fingerprint. This paper proposes some research directions to gather statistical knowledge of the within-source and between-sources variability of configurations of three minutiae on fingermarks and fingerprints. This paper proposes the use of the likelihood ratio (LR) approach to assess the value of fingerprint evidence. The model explores the statistical contribution of configurations of three minutiae using Tippett plots and related measures to assess the quality of the system. Features vectors used for statistical analysis have been obtained following a preprocessing step based on Gabor filtering and image processing to extract minutia position, type, and direction. Spatial relationships have been coded using Delaunay triangulation. The metric, used to assess similarity between two feature vectors is based on an Euclidean distance measure. The within-source variability has been estimated using a sample of 216 fingerprints from four fingers (two donors). Between-sources variability takes advantage of a database of 818 ulnar loops from randomly selected males. The results show that the data-driven approach adopted here is robust. The magnitude of LRs obtained under the prosecution and defense propositions stresses upon the major evidential contribution that small portions of fingermark, containing three minutiae, can provide regardless of its position on the general pattern.

This paper presents a statistical model for the quantification of the weight of fingerprint evidence. Contrarily to previous models (generative and score-based models), our model proposes to estimate the probability distributions of spatial relationships, directions and types of minutiae observed on fingerprints for any given fingermark. Our model is relying on an AFIS algorithm provided by 3M Cogent and on a dataset of more than 4,000,000 fingerprints to represent a sample from a relevant population of potential sources. The performance of our model was tested using several hundreds of minutiae configurations observed on a set of 565 fingermarks. In particular, the effects of various sub-populations of fingers (i.e., finger number, finger general pattern) on the expected evidential value of our test configurations were investigated. The performance of our model indicates that the spatial relationship between minutiae carries more evidential weight than their type or direction. Our results also indicate that the AFIS component of our model directly enables us to assign weight to fingerprint evidence without the need for the additional layer of complex statistical modeling involved by the estimation of the probability distributions of fingerprint features. In fact, it seems that the AFIS component is more sensitive to the sub-population effects than the other components of the model. Overall, the data generated during this research project contributes to support the idea that fingerprint evidence is a valuable forensic tool for the identification of individuals.

We respond to the article “Experimental results of fingerprint comparison validity and reliability: A review and critical analysis” by Ralph and Lyn Haber which offers (a) a one-sided criticism of the state of affairs in latent print examination, (b) lack of original data supporting that their suggested approach, and (c) a host of incorrect statements, inaccuracies, or obscure interpretations of the existing data.

We have a developed a method to apply a likelihood ratio (LR) statistic to ambiguous DNA profiles. These are DNA profiles that exhibit the phenomena of allele drop out (missing alleles) and allele drop in (contamination from a laboratory source). We have applied the theory to complex cases and tested the robustness of the model by using Tippett plots. The robustness of a model can be determined by the measurement of two errors: the chance that a truly innocent individual will give a LR greater than one (Type I error) and the chance that a truly guilty individual will give a LR less than one (Type II error).

Forensic examinations of ink have been performed since the beginning of the 20th century. Since the 1960s, the International Ink Library, maintained by the United States Secret Service, has supported those analyses. Until 2009, the search and identification of inks were essentially performed manually. This paper describes the results of a project designed to improve ink samples' analytical and search processes. The project focused on the development of improved standardization procedures to ensure the best possible reproducibility between analyses run on different HPTLC plates. The successful implementation of this new calibration method enabled the development of mathematical algorithms and of a software package to complement the existing ink library.

The ASTM standards on Writing Ink Identification (ASTM 1789-04) and on Writing Ink Comparison (ASTM 1422-05) are the most up-to-date guidelines that have been published on the forensic analysis of ink. The aim of these documents is to cover most aspects of the forensic analysis of ink evidence, from the analysis of ink samples, the comparison of the analytical profile of these samples (with the aim to differentiate them or not), through to the interpretation of the result of the examination of these samples in a forensic context. Significant evolutions in the technology available to forensic scientists, in the quality assurance requirements brought onto them, and in the understanding of frameworks to interpret forensic evidence have been made in recent years. This article reviews the two standards in the light of these evolutions and proposes some practical improvements in terms of the standardization of the analyses, the comparison of ink samples, and the interpretation of ink examination. Some of these suggestions have already been included in a DHS funded project aimed at creating a digital ink library for the United States Secret Service.

This article focuses on the interpretation of the outcome of forensic ink analysis. Forensic analysis of ink can provide valuable information to forensic investigation of questioned documents. Ink analysis can be interpreted in a comparison process or in an identification process. This article formalizes these two processes and discusses the current limitations of the interpretation of ink analysis. Keywords: ink; source; comparison; identification; interpretation

In the first part of this research, three stages were stated for a program to increase the information extracted from ink evidence and maximise its usefulness to the criminal and civil justice system. These stages are (a) develop a standard methodology for analysing ink samples by high-performance thin layer chromatography (HPTLC) in reproducible way, when ink samples are analysed at different time, locations and by different examiners; (b) compare automatically and objectively ink samples; and (c) define and evaluate theoretical framework for the use of ink evidence in forensic context. This report focuses on the second of the three stages. Using the calibration and acquisition process described in the previous report, mathematical algorithms are proposed to automatically and objectively compare ink samples. The performances of these algorithms are systematically studied for various chemical and forensic conditions using standard performance tests commonly used in biometrics studies. The results show that different algorithms are best suited for different tasks. Finally, this report demonstrates how modern analytical and computer technology can be used in the field of ink examination and how tools developed and successfully applied in other fields of forensic science can help maximising its impact within the field of questioned documents.