AI-Enabled Cybercrime: Exploring Adoption Trends in Underground Communities

• In Event: Cybercrime Unmasked: Myths, AI, and the Reality of Impunity

Abstract

The rapid advancement of artificial intelligence (AI) is predicted to disrupt the cybercrime landscape, offering new tools that will enhance the capabilities of low-level offenders. This presentation seeks to examine the adoption and use of AI technologies by cyber-offenders, with a focus on drivers and barriers, as well as on the products and services they seem to favor to automate attacks, generate malicious code, and craft effective social engineering scripts. The main research question is to what extent the accessibility of generative AI models and machine learning algorithms lowers barriers to entry for cyber-offenders. This study draws on a unique dataset of 250 conversations collected from a dozen cybercrime forums, including higher-level forums such as Exploit and XSS. AI-related topics are analyzed through a thematic analysis framework. Key themes identified in the dataset include curiosity about the potential of AI to enhance cybercrime skills, discussions on the performance of legitimate and malicious LLMs, innovative ideas about potential uses, the downsides of LLMs for the cybercrime community, and skepticism about the true value of AI tools for cybercrime uses. The methodology integrates digital ethnography with content analysis, providing an in-depth understanding of how AI is being weaponized by cybercriminals. The presentation will conclude with suggestions for law enforcement, regulators, policymakers, and cybersecurity professionals seeking to address this growing challenge.

Author

• Benoit Dupont, Université de Montréal