Forensic individualisation from biometric data.

This paper deals with the question of inference of identity of source in forensic science and focuses on the individualisation process of human beings from biometric data. Its goal is to explicitly describe the logical mechanisms that underpin the inference of identity of source in forensic science and to propose a concrete methodology for the forensic individualisation from biometric data. It also analyses the limits of several biometric modalities (fingerprint, face, voice) used for forensic individualisation. Firstly, the concept of identity and its application in forensic science is described. Secondly, the confusion between numerical and qualitative identity is addressed. Thirdly, the hypotheticaldeductive method is described as a logical approach for forensic individualisation and the likelihood ratio approach premised on the Bayes Theorem is proposed as a concrete methodology to infer identity of human beings from biometric data in forensic science. Finally a concrete framework is proposed, which includes hypotheses definition, sources and databases selection, analysis and comparison of biometric data and their interpretation.