ForensicBiometrics:
From twocommunities to OneDiscipline
Didier Meuwly*, Raymond Veldhuis**
*Netherlands Forensic Institute, WISK, 2490AAThe Hague,The Netherlands **The University of Twente,SAS,7500AEEnschede, TheNetherlands *[email protected] **[email protected]
Abstract: This articledescribes how thefieldsofbiometricsand forensic science can contributeand benefitfromeachother.The aimistofosterthe developmentof newmethods andtoolsimproving thecurrent forensic biometricapplications and allowing forthe creation of newones. Thearticle begins with adefinitionand a summary of thedevelopmentinforensicbiometrics. Then it describesthe data and biometricmodalities of interest in forensic scienceand theforensicapplications embedding biometrictechnology. On this basisitdescribes thesolutions and limitations of thecurrent practice regardingthe data,the technology andthe inferencemodels.Finally,itproposes research orientations forthe improvement of thecurrent forensic biometricapplications andsuggestssomeideas forthe developmentofsomenew forensic biometricapplications
1Introduction
Forensic scienceisdefined as thebody of scientific knowledge andtechnicalmethods used to analyseand interprettraces,inorder to answer questions relatedtocriminal, civil andadministrativelaw.Itfocuses in particular on thedemonstrationofthe existenceand theinvestigationofanoffence, on theindividualization of aperpetrator andonthe descriptionofamodus operandi.The practice of forensic scienceisfoundedon4basic inferences:identification, individualization, associationand reconstruction [IR98].These inferences arestructuredin3levels: thesource level, theactivitylevel andthe offence level[Co98].The source levelfocuses on thequestionofthe origin of atrace,the activity levelconcentrates on theactivitythatleads to atrace andthe offencelevel adresses thequestionifanactivityisconstitutiveofanoffence.
Biometricsisthe setofautomated methods used forthe recognition of human beings, measuringand analyzingstatistically theirdistinctivephysical andbehavioural traits. Themethod consists of theextractionand comparison of biometricfeaturesfroma referenceand atestsample, followedbythe computationofascorerepresentinga distance or asimilaritybetween thetwo samples[Wh10].
Currently scores areusedin3typesofforensicinferencesatsource level: identification andidentityverification, individualization, andassociation. Identificationand identity