Analyzing the DarkNetMarkets subreddit for evolutions of tools and trends using LDA topic modeling. (July 2018)

• Record Type: Journal Article
• Authors: Porter, Kyle
• Is Part Of: Digital investigation. Volume 26(2018)Supplement
• Page Start: S87
• Page End: S97
• Publication Date: 2018-07
• Subjects: Topic modeling -- Latent dirichlet allocation -- Web crawling -- Datamining -- Semantic analysis -- Digital forensics -- Surface-web monitoring; Forensic sciences -- Data processing -- Periodicals; Criminal investigation -- Data processing -- Periodicals; 363.250285
• DOI: 10.1016/j.diin.2018.04.023
• Languages: English
• ISSNs: 1742-2876

Abstract

Abstract: Darknet markets, which can be considered as online black markets, in general sell illegal items such as drugs, firearms, and malware. In July 2017, significant law enforcement operations compromised or completely took down multiple international darknet markets. To quickly understand how this affected the markets and the choice of tools utilized by users of darknet markets, we use unsupervised topic modeling techniques on the DarkNetMarkets subreddit in order to determine prominent topics and terms, and how they have changed over a year's time. After extracting, filtering out irrelevant posts, and preprocessing the text crawled from the subreddit, we perform Latent Dirichlet Allocation (LDA) topic modeling on a corpus of posts for each month from November 5th, 2016 to November 5th, 2017. Our results indicate that even analyzing public forums such as the DarkNetMarkets subreddit can reveal trends and keywords related to criminal activity and their methods, and that users of the dark web appear to be becoming increasingly more security-minded due to the recent law enforcement events.