TLDR: CYBERA has been selected to present its innovative AI scambaiting research, ‘Send to Which Account? Evaluation of an LLM-Based Scambaiting System,’ at the APWG eCrime 2025 Conference. This research introduces a novel method of engaging scammers using large language model (LLM)-driven conversational honeypots to gather crucial threat intelligence, demonstrating a proactive defense against AI-driven financial fraud.
CAMBRIDGE, Mass. – September 29, 2025 – CYBERA, a prominent entity in scam prevention and mule account intelligence, has announced its selection to present groundbreaking AI scambaiting research at the APWG eCrime 2025 Conference. The paper, titled ‘Send to Which Account? Evaluation of an LLM-Based Scambaiting System,’ will be showcased at the peer-reviewed symposium in San Diego this November.
The research addresses the escalating threat posed by scammers leveraging generative AI (GenAI) technologies to produce highly convincing phishing content at an unprecedented scale, thereby fueling financial fraud and eroding public trust. CYBERA’s innovative approach involves directly engaging these scammers through sophisticated conversational honeypots, powered by large language models (LLMs), to extract actionable threat intelligence.
During a rigorous five-month deployment, CYBERA’s system initiated over 2,600 conversations with active scammers, resulting in the generation of more than 18,700 messages. The system achieved a notable Information Disclosure Rate (IDR) of 32%, successfully capturing sensitive financial details, including mule bank accounts and cryptocurrency wallets. This intelligence is deemed critical for financial institutions, payment processors, and law enforcement agencies, enabling them to freeze illicit funds and protect consumers before financial losses occur.
Patrick Peterson, CEO of CYBERA, emphasized the company’s vision, stating, “At CYBERA, we believe that AI should not only protect people from scams but actively disrupt the criminal infrastructure that enables fraud. Our LLM-based scambaiting system represents a new frontier in proactive fraud defense – one where we can gather intelligence at scale and share it with financial institutions to stop scams before victims lose money.”
Members of the distinguished research team include Hossein Siadati from CYBERA, Haadi Jafarian from the University of Colorado Denver, and Sima Jafaikhah from UNCW. Their collaborative work not only highlights the potential of AI-driven engagement in enhancing fraud prevention but also identifies operational challenges, particularly the need to improve initial scammer engagement rates, offering valuable guidance for future advancements in this field.
Also Read:
- Santander UK Leverages AI to Disrupt Human Trafficking Networks
- AI Agent Immune System Revolutionizes Cybersecurity with 3.4x Faster Threat Containment and Minimal Overhead
The APWG eCrime 2025 Conference, marking its 20th annual peer-reviewed symposium, is set to examine critical factors in managing the global cybercrime plexus. This year’s conference will specifically focus on the role of AI and the growing menace of cyber-physical threats, with a theme titled ‘Cybercrimes Only AI and Crimebots Can Dream Of’. The symposium will explore AI’s impact as both a criminal co-conspirator and a defensive collaborator, including malicious AI agents for enhanced malware polymorphism and agentic spearphishing. The inclusion of CYBERA’s research underscores the conference’s commitment to addressing cutting-edge challenges and solutions in the evolving landscape of cybercrime.
