Alongside the numerous lawful activities on the Internet, several are linked to criminal and terrorist purposes. This is the terrain of the dark web and AI can reveal its usefulness, for criminals and the police
Employ techniques ofartificial intelligence in the dark web it’s an option both law enforcement and cyber offenders know about. AI makes it possible to carry out activities that are valuable to both parties. Just in the days in which the Privacy Guarantor decided to block ChatGPT for illicit collection of personal data, the European Police Office (Europol), published a report dedicated to the impact of ChatGPT and the wider family of linguistic models of Large Size (LLM) on Law Enforcement. Their misuse can also foreshadow the proliferation of “dark LLMs”, hosted on the dark web to provide unprotected chatbots, as well as trained on potentially harmful data.
Today there are over 1.5 billion sites on the World Wide Web. They seem very numerous, yet they are only the tip of a gigantic iceberg composed of a submerged part, called deep web of which the dark web. Before explaining what the deep and dark web are, it is good to clarify how much the world of crime has found in the “dark web” an ideal space for its illicit business. An example is DarkMarket, the world‘s largest illegal market on the dark web, vanquished by law enforcement in a joint international investigation that took offline a space in which almost 500,000 users and more than 2,400 sellers gravitated, where more than 320,000 transactions were carried out and over 4,650 bitcoins and 12,800 moneros were transferred, two of the most important cryptocurrencies in circulation. In this market the vendors traded drugs, smuggled counterfeit money, stolen or counterfeit credit cards, anonymous SIM cards and malware.
The desire to make money illegally and the possibility of doing so through the use of data fuel the curiosity and ambitions of cyber criminals towards artificial intelligence, as well as the interest of the police in using it to tackle the underworld and defeat it.
Deep web and dark web: a dark but mostly legitimate world
Before explaining the role of artificial intelligence in the dark web, it is good to explain what the surface and submerged web are. We have proposed the similarity of the Internet as an iceberg with only the tip visible: indexed sites are part of it. However, however voluminous the number of websites may appear in the open, the vast majority (quantifiable in about 95% of the total) is the submerged part, called the deep web.
«Every user can access the emerged part through a search engine. This does not happen with the deep web, made up mostly of websites of a legal nature in which information circulates, even useful and reserved for a specific community and for this reason they are not accessible to everyone», explains Nicola Ciani, researcher of the Digital Innovation Observatories of the Milan Polytechnic. In principle, anyone can access the deep web, but much of the content is protected.
The dark web is part of the deep web. It is based on a different network infrastructure, also called darknet, which includes – it is estimated – a few tens of thousands of websites. These are activities that are carried out in various countries. Not all of them are illegal: the dark web is also used by characters threatened by totalitarian regimes to communicate in a relatively safe way.
One of the most used software to access the dark web is Tor, an acronym for The onion router, free software platform. It works using a technique known as onion routing to establish anonymous communications on the public Internet. It uses an open source browser that routes data through an encrypted communications stack that resembles the layers of an onion; as it is structured, it allows you to mask the user’s identity through multiple encrypted passages between servers. Each node in the path can only know the one before and after.
Tor is called the largest anonymity network in the world. According to a report by Britannica, an average of 2.6 million users connect each day.
The evolution of the dark web and the role of AI
«Today we are witnessing a slightly different dark web, compared to the original one. Alongside activities such as the sale of weapons, drugs and child pornography, the sale of data has emerged. The current nature of the dark web is taking on that of a marketplace of stolen data and hacking services, or services aimed at attacking the activities of individuals and companies», explains Ciani.
What role do artificial intelligence techniques play in this “dark world“? Ambivalent. AI is used both by the police to sift through the darknet and identify criminal sites and activities, but also by the criminals themselves.
«There are at least two activities conducted by artificial intelligence on the dark web. The first is the monitoring: everything that is done for this purpose on the networks passes through AI algorithms that monitor a continuous and extremely large flow of data and the activities conducted on the web. The most adopted technique is machine learning, even if it is not the only one. The second activity, carried out in the last 3-4 years, is aimed at search for data on the dark webin particular those stolen and against their illicit circulation”.
Connecting users on the dark web is what law enforcement agencies are already trying to do. The problem is that the amount of data they need to manually scroll through is too large and unstructured for them to find connections quickly. Therefore, only a small percentage of cases can be prosecuted. To automate the process of linking people, MIT’s Lincoln Laboratory has trained machine learning algorithms to calculate the similarity between users on different forums and come to identify them. The Lab itself contributed to the Defense Advanced Research Projects Agency (DARPA) Memex program. Memex, which has produced a suite of software for data analysis on the dark web.
Artificial intelligence in the dark web: the reasons for the interest of cybercrime
The reason why criminals are interested in the use of artificial intelligence on the dark web is primarily economics. The turnover generated in the dark space of the web is difficult, but its magnitude can be understood. According to a recent research report by the analysis firm Chainalysis, the darknet market has reached 2.1 billion dollars in cryptocurrencies, part of which generated by fraudulent activities such as the sale of stolen logins, credit cards or other. These estimates tend to fall short: if we consider that the German police estimated that the largest money laundering service in the world on the darknet linked to the ChipMixer server was equal to 2.8 billion euros worth of bitcoins, we understand what could be revenue generated from data traffic. Then there is the drug trade and other lucrative illicit activities.
Then there is terrorist activity, which is also interested in the criminal use of technological innovation, including artificial intelligence, on the dark web. In February, specialist counter-terrorism units from 17 countries worked closely with Europol’s European Counter-Terrorism Center to restrict access to online instructions on how to use high-risk chemicals for terrorist attacks.
The ChatGPT dilemma and the possible use of artificial intelligence in the dark web
The use of artificial intelligence on the dark web is more topical than ever. The Independent recently highlighted how cybercrime is placing strong interest on generative AI, in particular on ChatGPT. The number of new posts dedicated to software developed by OpenAI on the dark web grew sevenfold between January and February, while the popularity of threads increased by 145%.
A separate report from cybersecurity firm Norton warned that criminals could be drawn to the advanced chatbot for its ability to produce responses that are mostly indistinguishable to a human.
Another aspect highlighted by Europol is that in addition to generating human-like language, ChatGPT is capable of producing code in different programming languages. “For a potential criminal with little technical knowledge, this is an invaluable resource for producing malicious code,” they write in the dedicated report.
In general, is AI also widespread for other illegal activities? «I think it is a tool used mostly in cyber attack activities, with the aim of obtaining the raw material to be exchanged on the darkweb», notes the researcher from the Milan Polytechnic.
breaking latest news © (Article protected by copyright)