CRII: SaTC: Towards Sustainable Intrusion Detection for Edge via Spatial Collaboration and Temporal Knowledge Retention

With the rapid growth of edge computing in critical sectors such as healthcare, smart cities, and distributed energy resource systems, securing these systems against cyber threats has become a national priority. However, conventional cybersecurity solutions struggle to meet the unique demands of edge environments, which are characterized by decentralized architectures, limited computing resources, and ever-changing threat patterns. This project aims to strengthen national cybersecurity by developing effective, adaptive, and collaborative Intrusion Detection Systems (IDS) specifically designed for edge environments. By enabling edge devices to share threat knowledge securely and retain critical information about past attacks, the research contributes to more resilient and sustainable network protection. The outcomes will advance scientific knowledge in cybersecurity and artificial intelligence, support public safety and national defense, and provide educational and outreach opportunities to equip the next generation of engineers and cybersecurity professionals. This project investigates two major challenges in deploying effective AI-based intrusion detection in edge computing environments: (1) how to enable decentralized, collaborative threat detection via spatial knowledge sharing among edge devices, and (2) how to ensure sustainable IDS performance through temporal knowledge retention. The proposed research introduces a federated learning-based framework that allows edge devices to train and share local IDS models without exposing raw data, aiming to preserve privacy and minimize communication overhead. A novel model aggregation strategy will prioritize contributions from devices that detect rare or novel threats using instability-based metrics. Meanwhile, the project develops temporal knowledge retention techniques, including prompt model patching and adaptive lifelong learning approaches, to ensure IDS models evolve with new threats while retaining past knowledge. A testbed implementation will validate the effectiveness and scalability of the proposed solutions in realistic scenarios. The results will contribute to the development of efficient, adaptive, and privacy-preserving IDS for edge computing and broader distributed systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.