The cheaters are caught at their own game! Researchers have developed Apate, an AI-powered chatbot that can speak out loud in natural language. His goal: to keep scammers online for as long as possible. Well seen !
Scams are a real nuisance, especially when it comes to phone scams, because you have to manage to outrun the scammers – unlike SMS and emails which you only have to delete in two seconds – and that costs us valuable time . It is enough that after a successful hack or phishing our phone number ends up on the dark web and we are harassed. GDPR or Banque de France scams, Personal Training Account (CPF) advertising, vishing… Cybercriminals have no lack of imagination to manipulate us!
Phone fraud is an extremely lucrative business that is growing every year. The ACCC (Australian Competition & Consumer Commission) estimates that Australians lost over €3.1 billion to scammers in 2022. Because even if phone providers block millions of fraudulent calls today, users are still being inundated with scams. And it doesn’t get any better… What if you had to catch cheaters at their own game to stop cheaters from cheating? This is, to say the least, the original idea that researchers at Macquarie University in Australia had. Led by Dali Kaafar, professor and executive director of the institution’s cybersecurity center, they decided to waste their time with scammers by keeping them online with a chatbot called Apate – a nod to the Greek goddess of deception. , infidelity, fraud, deceit, deceit and dishonesty.
Apate: an AI to waste crooks’ time
The idea for such a chatbot came to Dali Kaafar when he received a call from a scammer while having dinner with his family. Instead of hanging up, he decided to chat with him to make his kids laugh. Result: They stayed on the phone for 40 minutes. 40 minutes, time consuming to say the least, during which the scammer could not find any victims. “I realized that while I was wasting the scammer’s time so he couldn’t reach the vulnerable, and that was the point, it was also 40 minutes of my own life that I hadn’t restored,” explains Professor Kaafar the Macquarie University website. “Then I started thinking about how we could automate the whole process and use natural language processing to develop a computer-assisted chatbot that could have a credible conversation with the scammer.” Thus Apate was born. “We are very pleased that this new technology can destroy the business model of fraudulent calls and make them unprofitable,” says the researcher.
To develop Apate, the Macquarie University Cybersecurity Center team first analyzed fraudulent phone calls and identified the social engineering techniques scammers use on their victims – to create a sense of urgency, impersonate authority, threaten punishment, etc – Using machine learning and natural language processing techniques to identify typical fraud scenarios. To do this, the researchers used a dataset of real-world fraudulent conversations, including call recordings, email transcripts, and social media chats, to allow the chatbot to generate its own conversations in a fraudulent situation. Professor Kaafar says he was surprised by Apate’s adaptability, which would have responded well to situations where she hadn’t been trained.
The researcher explains that thanks to advances in natural language processing (NLP) and artificial intelligence cloning of human voices, it is now possible to develop AI-powered conversational agents capable of speaking in natural language, accepting a specific person and maintaining an engaging conversation through consistent responses. “The chatbots we’ve developed can trick scammers into thinking they’re talking to potential fraud victims, so they spend time cheating on the bots,” says the professor. Kaafar. The technology, for which a patent has already been filed, is still under development. Apate initially only occupies scammers for 5 minutes instead of the targeted 40 minutes. It could be of interest to operators – some of whom the researchers are also in talks with – but also to various institutions (governments, banks, etc.) by collecting a lot of data on the development of scams.