Shedding New Light on the Language of the Dark Web

Abstract

The hidden nature and the limited accessibility of the Dark Web, combined with the lack of public datasets in this domain, make it difficult to study its inherent characteristics such as linguistic properties. Previous works on text classification of Dark Web domain have suggested that the use of deep neural models may be ineffective, potentially due to the linguistic differences between the Dark and Surface Webs. However, not much work has been done to uncover the linguistic characteristics of the Dark Web. This paper introduces CoDA, a publicly available Dark Web dataset consisting of 10000 web documents tailored towards text-based Dark Web analysis. By leveraging CoDA, we conduct a thorough linguistic analysis of the Dark Web and examine the textual differences between the Dark Web and the Surface Web. We also assess the performance of various methods of Dark Web page classification. Finally, we compare CoDA with an existing public Dark Web dataset and evaluate their suitability for various use cases.

VendorLink: An NLP approach for Identifying & Linking Vendor Migrants & Potential Aliases on Darknet Markets

“…With the advances in NLP, there has been considerable research into the field of AA that has demonstrated the success of TFIDF-based clustering and classification techniques (Agarwal et al, 2019;İzzet Bozkurt et al, 2007), CNNs (Rhodes, 2015;Shrestha et al, 2017), RNNs (Zhao et al, 2018;Jafariakinabad et al, 2019;Gupta et al, 2019), and transformers architectures (Fabien et al, 2020;Ordoñez et al, 2020;Uchendu et al, 2020a). Moreover, researchers have also observed a significant difference in language structure between Darknet and Surface net websites (Choshen et al, 2019;Jin et al, 2022). Therefore, exploring the application of authorship tasks on the Darknet language is crucial.…”

“…To aid LEA, we perform supervised pre-training by conducting multiclass classification in a closed-set environment (Zhou et al, 2021a) to analyze different writing styles in text ads and classify vendor migrants to unique vendor accounts across three Darknet markets. Moreover, researchers have observed a significant difference in language structure between Darknet and Surface net websites (Choshen et al, 2019;Jin et al, 2022). Since most contextualized models are trained on surface web data, the supervised pre-training step allows our model to adapt to the Darknet market domain knowledge.…”