This research introduces a novel foundation model approach for detecting complex events in CPS-IoT systems, enabling more sophisticated security monitoring and decision-making.

• Addresses the limitation of existing models that focus only on short-span perception tasks
• Develops methods for long-term reasoning across sequences of atomic events
• Proposes spatiotemporal dependency modeling to filter irrelevant data
• Demonstrates effectiveness in smart monitoring and security applications

This advancement represents a significant step toward more robust security monitoring in critical infrastructure, smart factories, and autonomous systems where detecting complex patterns is essential for threat prevention.

Toward Foundation Models for Online Complex Event Detection in CPS-IoT: A Case Study