Forensic Investigations in the Age of AI: Identifying and Analyzing Artifacts from AI-Assisted Crimes

This paper examines the possible manipulation of generative AI models, in particular ChatGPT, to support cybercriminal activities. The research investigates the possibility of coercing AI into assisting malicious actors by setting up a suspect in a cybercrime scenario. In this paper, using a Windows machine and an Android Mobile, we simulate a scenario in which ChatGPT is manipulated to assist a cybercriminal and conduct forensic analysis to reveal evidence of AI involvement. Through integrated memory, mobile, and network forensics, we successfully extract AI-generated artifacts including session tokens, interaction histories, and manipulated prompts that circumvent AI safety measures. Key findings demonstrate how forensic techniques can effectively identify AI involvement in cybercrimes through volatile memory analysis and network traffic examination and how prompt engineering can successfully bypass AI safety measures, exposing vulnerabilities in current safeguards. The paper highlights both opportunities and risks for generative AI in criminal contexts through experimental analysis and forensic techniques. It further discusses ethical and legal issues concerning accountability, misuse of technology, and the potential for harmful behaviors facilitated by AI. These findings demonstrate an imperative need for robust safeguards and specialized forensic approaches to address the exploitation of AI in criminal activities.