Bayesian analysis of fingerprint, face and signature evidences with automatic biometric systems

The Bayesian approach provides a unified and logical framework for the analysis of evidence and to provide results in the form of likelihood ratios (LR) from the forensic laboratory to court. In this contribution we want to clarify how the biometric scientist or laboratory can adapt their conventional biometric systems or technologies to work according to this Bayesian approach. Forensic systems providing their results in the form of LR will be assessed through Tippett plots, which give a clear representation of the LR-based performance both for targets (the suspect is the author/source of the test pattern) and non-targets. However, the computation procedures of the LR values, especially with biometric evidences, are still an open issue. Reliable estimation techniques showing good generalization properties for the estimation of the between- and within-source variabilities of the test pattern are required, as variance restriction techniques in the within-source density estimation to stand for the variability of the source with the course of time. Fingerprint, face and on-line signature recognition systems will be adapted to work according to this Bayesian approach showing both the likelihood ratios range in each application and the adequacy of these biometric techniques to the daily forensic work.     

Computation of likelihood ratios in fingerprint identification for configurations of any number of minutiae

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.     

Computation of likelihood ratios in fingerprint identification for configurations of three minutiae

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.     

Using the likelihood ratio in bloodstain pattern analysis

There is an apparent paradox that the likelihood ratio (LR) approach is an appropriate measure of the weight of evidence when forensic findings have to be evaluated in court, while it is typically not used by bloodstain pattern analysis (BPA) experts. This commentary evaluates how the scope and methods of BPA relate to several types of evaluative propositions and methods to which LRs are applicable. As a result of this evaluation, we show how specificities in scope (BPA being about activities rather than source identification), gaps in the underlying science base, and the reliance on a wide range of methods render the use of LRs in BPA more complex than in some other forensic disciplines. Three directions are identified for BPA research and training, which would facilitate and widen the use of LRs: research in the underlying physics; the development of a culture of data sharing; and the development of training material on the required statistical background. An example of how recent fluid dynamics research in BPA can lead to the use of LR is provided. We conclude that an LR framework is fully applicable to BPA, provided methodic efforts and significant developments occur along the three outlined directions.     

Reporting likelihood ratio for casework in firearm evidence identification

Likelihood ratio (LR) plays an important role in estimating the weight of evidence in firearm evidence identifications. LR is computed from a statistical model including the distribution of the known-matching (KM or within) and known-nonmatching (KNM or between) comparison scores. Current LR procedures rely on KM/KNM scores from existing reference firearm toolmark data sets or alternatively from generating a set of test fires using multiple firearms. Both procedures may contain theoretical or practical issues which may hinder the LR procedures from reporting an unbiased LR estimation in casework. In this paper, a reference data set was established from a set of firearms, each test-fired two cartridge cases, resulting in a basic data set and a control data set. The congruent matching cells (CMC) method was used to generate CMC scores that are used to fit in the KM/KNM statistical distributions for LR estimation. In the initial test, 130 firearms from eight manufacturers were used for generating a reference data set consisting of 260 cartridge cases representing 130 KM and 8385 KNM pairwise breech face images. Test results showed that the KM and KNM distribution intersect at CMC = 2, which is equivalent to LR = 1 (equally to support both the prosecutor and the defense propositions). When the CMC threshold is increased to 6 or more, the LR values are higher than a million, which can provide extremely strong support to the conclusion of the same firearm (or the prosecutor's proposition) in the casework of firearm evidence identification.     

A Two‐Level Model for Evidence Evaluation in the Presence of Zeros*

Abstract: Likelihood ratios (LRs) provide a natural way of computing the value of evidence under competing propositions. We propose LR models for classification and comparison that extend the ideas of Aitken, Zadora, and Lucy and Aitken and Lucy to include consideration of zeros. Instead of substituting zeros by a small value, we view the presence of zeros as informative and model it using Bernoulli distributions. The proposed models are used for evaluation of forensic glass (comparison and classification problem) and paint data (comparison problem). Two hundred and sixty‐four glass samples were analyzed by scanning electron microscopy, coupled with an energy dispersive X‐ray spectrometer method and 36 acrylic topcoat paint samples by pyrolysis gas chromatography hyphened with mass spectrometer method. The proposed LR model gave very satisfactory results for the glass comparison problem and for most of the classification tasks for glass. Results of comparison of paints were also highly satisfactory, with only 3.0% false positive answers and 2.8% false negative answers.     

Bayesian networks and the value of the evidence for the forensic two-trace transfer problem*

Forensic scientists face increasingly complex inference problems for evaluating likelihood ratios (LRs) for an appropriate pair of propositions. Up to now, scientists and statisticians have derived LR formulae using an algebraic approach. However, this approach reaches its limits when addressing cases with an increasing number of variables and dependence relationships between these variables. In this study, we suggest using a graphical approach, based on the construction of Bayesian networks (BNs). We first construct a BN that captures the problem, and then deduce the expression for calculating the LR from this model to compare it with existing LR formulae. We illustrate this idea by applying it to the evaluation of an activity level LR in the context of the two-trace transfer problem. Our approach allows us to relax assumptions made in previous LR developments, produce a new LR formula for the two-trace transfer problem and generalize this scenario to n traces.     

亲权指数和个体识别似然比计算软件的开发与应用

目的建立计算亲子鉴定亲权指数(PI)和个体识别似然比(LR)的计算软件。方法依据相关行业规范和文献中给出的计算方法,利用计算机语言Visual Basic 6.0编写程序。结果开发出适用于PI和LR的计算软件。结论该计算软件可以帮助工作人员提高计算效率,服务法医物证工作。     

Application of Random Match Probability Calculations to Mixed STR Profiles

Mixed DNA profiles are being encountered more frequently as laboratories analyze increasing amounts of touch evidence. If it is determined that an individual could be a possible contributor to the mixture, it is necessary to perform a statistical analysis to allow an assignment of weight to the evidence. Currently, the combined probability of inclusion (CPI) and the likelihood ratio (LR) are the most commonly used methods to perform the statistical analysis. A third method, random match probability (RMP), is available. This article compares the advantages and disadvantages of the CPI and LR methods to the RMP method. We demonstrate that although the LR method is still considered the most powerful of the binary methods, the RMP and LR methods make similar use of the observed data such as peak height, assumed number of contributors, and known contributors where the CPI calculation tends to waste information and be less informative.     

司法证明模糊论

“真实”本身具有多元解释,是一个不可能得到精确认定的模糊概念。“事实”必须为法官所确立,是一种个性思维过程后的结论,可称为“法官真实”。从模糊理论与心证演变模式的三个阶段的视角看,自由心证的实质是模糊心证。司法证明必须容忍甚至鼓励富有极大模糊性的日常生活语言以描绘案件事实。模糊理论为司法证明科学的发展带来了新的视角与方法,扩展了研究者理性选择的空间,做出了不同于既有法学语境的诠释,试图形成一个能够科学描述和处理司法证明模糊性的概念体系和方法论框架,以突破摇摆于确定性与不确定性之间的困惑。     

Weight of authorship evidence with multiple categories of stylometric features: A multinomial-based discrete model

This study empirically demonstrates the efficacy of a two-level Dirichlet-multinomial statistical model (the Multinomial system) for computing likelihood ratios (LR) for linguistic, textual evidence with multiple stylometric feature types with discrete values. The LRs are calculated separately for each feature type, namely, word, character and part of speech N-grams (N = 1,2,3), which are combined as overall LRs through logistic regression fusion. The Multinomial system’s performance is compared with that of a previously proposed system with the cosine distance (the Cosine system) using the same data (i.e., documents collated from 2160 authors). The experimental results show that: (1) the Multinomial system outperforms the Cosine system with the fused feature types by a log-LR cost of ca. 0.01 ～ 0.05 bits; and (2) the Multinomial system is more advantageous in performance with longer documents than the Cosine system. Although the Cosine system is more robust overall against the sampling variability arising from the number of authors included in the reference and calibration databases, the Multinomial system can achieve reasonable stability in performance; for example, the standard deviation value of the log-LR cost becomes lower than 0.01 (10 random samplings of authors for the reference and calibration databases) with 60 or more authors in each database.     

Classification of Glass Fragments Based on Elemental Composition and Refractive Index*

Abstract: The aim of this study was to assess the efficiency of likelihood ratio (LR)‐based measures when they are applied to solving various classification problems for glass objects which are described by elemental composition, and refractive index (RI) values, and compare LR‐based methods to other classification methods such as support vector machines (SVM) and naïve Bayes classifiers (NBC). One hundred and fifty‐three glass objects (23 building windows, 25 bulbs, 32 car windows, 57 containers, and 16 headlamps) were analyzed by scanning electron microscopy coupled with an energy dispersive X‐ray spectrometer. Refractive indices for building and car windows were measured before (RI_b), and after (RI_a) an annealing process. The proposed scheme for glass fragment(s) classification demonstrates some efficiency, although the classification of car windows (c) and building windows (w) must be treated carefully. This is because of their very similar elemental content. However, a combination of elemental content and information on the change in RI during annealing (ΔRI = RI_a?RI_b) gave very promising results. A LR model for the classification of glass fragments into use‐type categories for forensic purposes gives slightly higher misclassification rates than SVM and NBC. However, the observed differences between results obtained by all three approaches were very similar, especially when applied to the car window and building window classification problem. Therefore, the LR model can be recommended because of the ease of interpretation of LR‐based measures of certainty.     

Handwriting evidence evaluation based on the shape of characters: application of multivariate likelihood ratios

A novel Bayesian methodology has been developed to quantitatively assess handwriting evidence by means of a likelihood ratio (LR) designed for multivariate data. This methodology is presented and its applicability is shown through a simulated case of a threatening anonymous text where a suspect is apprehended. The shape of handwritten characters a, d, o, and q of the threatening text was compared with characters of the true writer, and then with two other writers, one with similar and one with dissimilar characters shape compared to the true writer. In each of these three situations, 100 draws of characters were made and the resulting distributions of LR were established to consider the natural handwriting variation. LR values supported the correct hypothesis in every case. This original Bayesian methodology provides a coherent and rigorous tool for the assessment of handwriting evidence, contributing undoubtedly to integrate the field of handwriting examination into science.     

DNA mixture genotyping by probabilistic computer interpretation of binomially-sampled laser captured cell populations: Combining quantitative data for greater identification information

Two person DNA admixtures are frequently encountered in criminal cases and their interpretation can be challenging, particularly if the amount of DNA contributed by both individuals is approximately equal. Due to an inevitable degree of uncertainty in the constituent genotypes, reduced statistical weight is given to the mixture evidence compared to that expected from the constituent single source contributors. The ultimate goal of mixture analysis, then, is to precisely discern the constituent genotypes and here we posit a novel strategy to accomplish this. We hypothesised that LCM-mediated isolation of multiple groups of cells (‘binomial sampling’) from the admixture would create separate cell sub-populations with differing constituent weight ratios. Furthermore we predicted that interpreting the resulting DNA profiling data by the quantitative computer-based TrueAllele® interpretation system would result in an efficient recovery of the constituent genotypes due to newfound abilities to compute a maximum LR from sub-samples with skewed weight ratios, and to jointly interpret all possible pairings of sub-samples using a joint likelihood function.As a proof of concept, 10 separate cell samplings of size 20 recovered by LCM from each of two 1:1 buccal cell mixtures were DNA-STR profiled using a specifically developed LCN methodology, with the data analyzed by the TrueAllele® Casework system. In accordance with the binomial sampling hypothesis, the sub-samples exhibited weight ratios that were well dispersed from the 50% center value (50 ± 35% at the 95% level). The maximum log(LR) information for a genotype inferred from a single 20 cell sample was 18.5 ban, with an average log(LR) information of 11.7 ban. Co-inferring genotypes using a joint likelihood function with two sub-samples essentially recovered the full genotype information. We demonstrate that a similar gain in genotype information can be obtained with standard (28-cycle) PCR conditions using the same joint interpretation methods. Finally, we discuss the implications of this work for routine forensic practice.     

网络证据的特殊性及研究价值

网络证据是指从不同网络终端获取的信息载体。它是电子证据中一类独特的存在形式。与其他形式的电子证据不同,网络证据可以通过不同地点、不同空间在网上获取。因此,网络证据的收集、审查、认定都有着特殊性。明确界定网络证据的概念、特征,对于更好地开发证据调查、收集的新方法,在司法实践中更好地查明案件事实;对于修改和完善我国立法关于诉讼证据及其运用规则的规定;对于我们从新的角度去理解证据及其相关基础理论问题,都有着极其重要的意义。     

The COLLECTORS ranking scale for ‘at‐scene’ digital device triage

As digital evidence now features prominently in many criminal investigations, such large volumes of requests for the forensic examination of devices has led to well publicized backlogs and delays. In an effort to cope, triage policies are frequently implemented in order to reduce the number of digital devices which are seized unnecessarily. Often first responders are tasked with performing triage at scene in order to decide whether any identified devices should be seized and submitted for forensic examination. In some cases, this is done with the assistance of software which allows device content to be “previewed”; however, in some cases, a first responder will triage devices using their judgment and experience alone, absent of knowledge of the devices content, referred to as “decision‐based device triage” (DBDT). This work provides a discussion of the challenges first responders face when carrying out DBDT at scene. In response, the COLLECTORS ranking scale is proposed to help first responders carry out DBDT and to formalize this process in an effort to support quality control of this practice. The COLLECTORS ranking scale consists of 10 categories which first responders should rank a given device against. Each devices cumulative score should be queried against the defined “seizure thresholds” which offer support to first responders in assessing when to seize a device. To offer clarify, an example use‐case involving the COLLECTORS ranking scale is included, highlighting its application when faced with multiple digital devices at scene.     

论法庭证据评估体系的发展

随着法律和科学的发展,人们对部分传统的法庭科学产生了质疑。本文探讨了传统法庭证据检验的弊端和目前国际上法庭科学证据的接纳原则,推出了似然比法庭证据检验评估体系,包括似然比对证据强度的量化方法、检验结论的科学解释方法和法庭识别系统可靠性的评估方法。