Online fraud has long been an issue of concern to forensic scientists, particularly in terms of detection of fraudulent ads. Indeed, detecting a fraudulent ad is not always obvious, especially when the victims themselves do not realise that they are victims of a fraud (Rossy and Ribaux, 2020). However, when these are detected, it becomes interesting to approach them from a forensic intelligence perspective in order to detect crime series. The detection of links between cases is based on traces of different nature (Ribaux et al., 2022). The objective of this paper is to determine whether it is possible to detect links between fraudulent ads using the textual data available, through the language trace (Renaut et al., 2017) and a methodology derived from forensic linguistics. This is a field of research involving statistical analysis of textual data or style analysis for various purposes in a judicial context, including author recognition. These approaches can be used to identify author-specific features, which can be assimilated to language "traces" that can be used in different contexts, such as authorship analysis or the detection of similarities between two texts.
To understand the issues faced by specialists in fraud detection, a preliminary study is made through interviews with online crime analysts. Since online frauds may happen in both legal and illegal online markets, data from three Dark Web cryptomarkets are analysed: Empire Market (data collected from April 2020 to August 2020), White House Market (data collected from April 2020 to March 2021), and Darkmarket (data collected from July 2020 to June 2021). By hypotheses, they indeed allow distinguishing frauds through the feedback left by buyers. The textual data from the ad description and the texts present in the ads evaluations are analysed. The first step consists of distinguishing legitimate ads from fraudulent ads based on feedback. Then, we detect patterns in the ads and infer links between the fraudulent ads. A non-exhaustive list of linguistic characteristics common to fraudulent ads is then formalised.
Renaut L, Ascone L and Longhi J (2017) De la trace langagière à l'indice linguistique : enjeux et précautions d'une linguistique forensique. Ela. Études de linguistique appliquée (188): 423–442.
Ribaux O, Baechler S and Rossy Q (2022) Forensic Intelligence and Traceology in Digitalised Environments: The Detection and Analysis of Crime Patterns to Inform Practice. In: Gill M (ed.) The Handbook of Security. Cham: Springer International Publishing, pp. 81–99. DOI: 10.1007/978-3-030-91735-7_5.
Rossy Q and Ribaux O (2020) Orienting the Development of Crime Analysis Processes in Police Organisations Covering the Digital Transformations of Fraud Mechanisms. European Journal on Criminal Policy and Research 26(3): 335–356. DOI: 10.1007/s10610-020-09438-3.
