Evidential value of duct tape comparison: A stochastic model based on loopbreaking patterns

Abstract

Pieces of duct tape are retrieved from many crime scenes. They are a great source for forensic traces. The evidential value of these traces depend on the position of the piece of duct tape with respect to the roll and potential other pieces of duct tape. This can be determined based on findings that two pieces of duct tape used to be connected to each other. Forensic experts at the NFI proposed a new, additional method of examining this, which uses loopbreaking patterns. We construct a likelihood ratio-system which evaluates the likelihood ratios of these loopbreaking patterns. This LR-system consists of dynamic Bayesian networks, which assumes that the loopbreaking patterns comply to the Markov property. A small data set was available for the training and testing of this system. The validation of this system is evaluated in terms of accuracy, discriminating power and calibration. The results of the validation seem promising. For future work we recommend to validate the system on data which represent real forensic cases.