This research introduces IoTGen, an innovative framework that leverages LLMs to generate synthetic smart home user behavior sequences, solving the data scarcity problem in security testing.

• Creates realistic, diverse user behavior patterns without privacy concerns
• Combines Structure Pattern Generator and LLM-based Content Generator for comprehensive data synthesis
• Outperforms existing methods with up to 93.2% realistic behavior generation
• Enables more robust anomaly detection and security testing for smart home environments

Why it matters: Smart home security solutions currently rely on limited, static datasets that fail to adapt to evolving threats and usage patterns. This approach allows continuous generation of high-quality training data without compromising user privacy.

Synthetic User Behavior Sequence Generation with Large Language Models for Smart Homes