APPLYING TOPIC MODELLING ON FORENSIC DATA: A CASE STUDY

Most actionable evidence for investigation purposes is identified during the analysis phase of a digital investigation process. The objective of the analysis phase (digital analysis) is to reduce the quantity and enhance the intelligibility of data that must be reviewed by a human analyst. Currently, this is done through expression based searching, which assumes a good understanding of the evidence prior to the search. Therefore, latent evidence will not be found with such methods. This suggests a clear role for knowledge discovery and data mining (KDD) techniques to enhance the digital analysis process. The research described in this article investigates the application of topic modelling as a KDD technique on forensic data and its ability to contribute to digital analysis. Topic models infer the underlying semantic context of a text collection and summarises it as topics described by words. The data used for this application was extracted from a real digital investigation case. This novel application highlights several challenges that forensic data poses to topic modelling algorithms and we report on lessons learned from the