Santi Romeu, head of the data science and analytics unit of the Cybersecurity Agency of Catalonia, explains that ChatGPT is limited so that it cannot be an accomplice to criminal activities, such as a scam for phishing. However, he also admits that it is possible to deceive him. It can be as easy as arguing that it’s for a report. Romeu accepts that AI has opened a new paradigm in the field of cybercrime. “Texts can be made massively, very well written, convincing and adapted to the recipients.” Before they had to be done manually. Now, thousands can be created with one click. And there are tools similar to ChatGPT but without limitations.
The turning point was in 2022, when ChatGPT went public. Since then, according to data from the Cybersecurity Agency of Catalonia, cyber security incidents related to the use of AI have increased by 370% worldwide. Romeu cites a study by Signicat, a company dedicated to digital identity solutions, from this September, which focuses on the field of cyber scams: 42.5% of cyber scams involve the use of AI to part of the attackers, and an estimated 29% success rate. In a report this summer, Europol already warns that the use of AI is beginning to be present in the day-to-day life of cybercriminals, and warns of its potential on the rise.
Generative AI allows you to generate texts, images, audios, videos… It can also be used to make a video of a celebrity saying that he has made a lot of money by investing in a way, or an audio impersonating the voice of a child asking help, or maintain an expert conversation in any language and on any subject. They have even impersonated businessmen in video conferences who have ordered bank transfers. It is known as technology deepfake.
February 2024, in Hong Kong: A worker of a multinational company paid $25 million, as agreed in a meeting with another company. The assistants, however, were deepfakes made with AI. June 2023, in the United States: A mother received a call. On the other side of the phone was his daughter saying that she had been kidnapped. But he wasn’t, his voice had been generated with AI.
According to Romeu, AIs can even create computer viruses. “Without knowing how to program, you can end up with a malicious program,” he points out, reiterating that this is available in generative AIs that are restricted in use. The world of cyber scams works on the system of crime as a servicethat is, jobs can be subcontracted. “An individual can buy one phishing“, say sources from the Mossos. Catalan police investigators point out that they have encountered multiple repeat offenders who spent their days committing thefts who are now dedicated to cyber scams because that way they don’t expose themselves as much.
Filtered data
Cybercriminals are leveraging AI to create personalized scams, and Romeu explains that fraudsters often use information obtained through cyberattacks and data leaks: “With all the data leaks out there, we all have something leaked on the secure internet,” he says. . This data allows them to customize the scams with artificial intelligence, which also serves them to make increasingly neat messages, in your language, flawless and perfectly believable. More and more convincing.
Artificial intelligence is constantly improving, and is also being used to fight crime. In the Agency they are using it to automate processes and deal with huge data sets that could not be processed without this technology.
