As people are using digital and online services more during the COVID-19 lockdown, there was also an increase in illegal activity in the darknet. As things are becoming more complicated, identifying vendor profiles in the darknet marketplaces can potentially help law enforcement agencies crack down on these illegal markets.
Using the dataset from the darknet archives, which is a collection of darknet marketplace HTML pages, we analyzed two of these markets (Silkroad 2 and Nucleus) totaling to around 50 GB of space. This was processed in AWS using Dask. Through K-Means clustering and association rule mining, we generated six clusters that were categorized into four general clusters.
We observed that although the platform is international, the transactions are highly localized. The results also showed interrelatedness in seemingly unrelated items, such as steroid items and opioids. Which is interesting, since there were already studies about their relationship in prior health studies.