Inducing Personality in LLM-Based Honeypot Agents: Measuring the Effect on Human-Like Agenda Generation
This work addresses the need for more realistic cyber deception tools to engage attackers, though it appears incremental in applying existing personality models to LLMs.
The paper tackled the problem of creating convincing cyber decoys by using Large Language Models (LLMs) to emulate human-like personalities, demonstrating that a prompt schema based on the five-factor model systematically induces distinct personalities in LLMs to improve cyber deception strategies.
This paper presents SANDMAN, an architecture for cyber deception that leverages Language Agents to emulate convincing human simulacra. Our 'Deceptive Agents' serve as advanced cyber decoys, designed for high-fidelity engagement with attackers by extending the observation period of attack behaviours. Through experimentation, measurement, and analysis, we demonstrate how a prompt schema based on the five-factor model of personality systematically induces distinct 'personalities' in Large Language Models. Our results highlight the feasibility of persona-driven Language Agents for generating diverse, realistic behaviours, ultimately improving cyber deception strategies.