DarkBERT is a large language model that was created to expand upon the capabilities of existing models such as BERT and RoBERTa. Research by a South Korean team suggests that DarkBERT may be equal or slightly superior to other language models of the same architecture type. This makes DarkBERT particularly useful in cases where it requires contextual understanding of a domain specific model like the Dark Web.
The team connected a language model to the dark web via Tor and collected raw data to create their DarkBERT model. The model was tested against existing AI models such as BERT and RoBERTa. It was found that DarkBERT exceeded them in terms of its contextual understanding of the Darknet, thereby outperforming both models.
It is not certain that DarkBERT will be available to the public or even to cybercriminals, but it may be used by law enforcement agencies who are looking to fight cybercrime but need the help of AI models to understand the language of the dark web. Journalists, opposition activists, and politically persecuted people also use the darknet to accesscensored content.
The Tor browser is the layer of secrecy that lets its users hide their online activity and data traffic from those who may want to analyse it. DarkBERT is an excellent resource for those who want to keep their online activity anonymous and still remain up to date with the most prevalent news on the darknet.
Meta AI/FAIR (Facebook AI Research) is the team behind DarkBERT and they offer the model to academics for research purposes. It remains to be seen whether the model will eventually be available to law enforcement agencies or the general public, but if leaked, DarkBERT could be used to better understand the darknet by offering valuable insights. As of now, due to its specialized approach, DarkBERT is still off-limits for most.
