Talk E-1: Beyond the Hype: Research on how Cybercriminals are Really Using GenAI

Cybercriminals were abusing AI well before the recent hype around generative AI took the IT industry by storm. This session delves into the criminal underground forums in order to determine how AI is actually being used, how they are adopting this technology to further their goals and what kind of AI-powered criminal services there are on offer. First, we will provide an overview of the underground conversations on AI and how the interest in generative AI there has followed the general market trends. We will highlight examples of forums dedicated to AI, such as Hack Forums’ "Dark AI" section. We will also cover LLM offerings by criminals for criminals with features and prices. As part of the previous, we have seen FraudGPT, DarkBARD, DarkBERT, DarkGPT sharing many features that make us believe that these most likely work as wrapper services to the legitimate ChatGPT or Google BARD. We also look at other possibly fake criminal LLM offerings: WolfGPT, XXXGPT, and Evil-GPT. We will share descriptions and speculation as to whether they are legitimate tools or also just wrappers of ChatGPT AI or OpenAI. Finally, the speakers will discuss deepfake services for criminals. While there has been a lot of hype around these services, we note that the creation process is labor-intensive and expensive, and it is still very difficult to create convincing deepfake videos. Audio deepfakes, however, are a lot easier to create – and we will demonstrate why and how these will become mainstream in the criminal market much sooner than video deepfakes. The session will conclude by summarizing the top three threats to worry about right now and three that may be of future concern. Time will be left for Q&A.