In South Korea, a group of researchers has managed to create a new linguistic model called DarkBERT, a sort of "controller" of the dark web to monitor illicit information.
What would happen if we programmed a machine learning algorithm to go down the rabbit hole, training it to dive into the dark web?
The answer is close, thanks to DarkBERT, a Large Language Model (LLM) trained in the dark web who, thanks to the RoBERTa architecture on which it is based, manages to learn a huge amount of information in a very short time, so as to develop the best form of communication to integrate into the dark web and find any hidden information.
The dark web is in fact a hidden part of the Internet that can only be accessed through special browsers like Tor. Its purpose is to keep your online activity anonymous and private, often in support of illegal actions and applications, but not limited to, because some users use it to bypass censorship imposed by their government.
Indeed, the anonymity of the dark web attracts many users who would be in danger if they revealed their identity online: victims of abuse, political dissidents and activists. However, the advantages of the dark web are also exploited by those who choose to act outside the law, with explicitly illegal methods.
Indeed, the dark web has billions of hidden web pages which include groups, blogs and shops where illegal actions are mainly carried out. Therefore, its study and surveillance are particularly complex even for the special security bodies.
The DarkBERT enhancement has required, and will require, constant updating over time; however, thanks to his help it will finally be possible to intercept the criminals who operate in the shadows (from hackers to contract hitmen).
Thanks to this tool and the use of the browser Tor, over 1.2 billion words have been searched which has made the AI increasingly performing and capable of generating content to further delve into the most hidden depths of the dark web.
The South Korean team claims the software is far more effective at "making sense of the dark web" than other models that have been trained to complete similar tasks.
Insiders suggest that DarkBERT can be used for a variety of cybersecurity-related tasks, such as detecting sites that sell ransomware or disclose confidential data.
Original article published on Money.it Italy 2023-05-25 08:00:00. Original title: Arriva DarkBERT, la nuova IA addestrata sul dark web
