Optimal Feature Selection for Intrusion Detection in Medical Cyber-Physical Systems

摘要

Medical cyber physical systems (MCPS) integrate the physical, communication and computation components of medical devices to enhance the quality and reliability of healthcare systems. With the remarkable progress of MCPS technologies in recent years, there is a need to advance the security measures to efficiently detect attacks in this domain. Research on intrusion detection for medical cyber physical systems is still in its infancy. For an efficient intrusion detection system (IDS), it is important to address the problem of feature selection to remove redundant, irrelevant and noisy features. Feature selection is even more relevant to address in MCPS as the use of entire feature space places unnecessary burden on resource constrained systems in this domain. Also since real-time detection of attacks is critical in healthcare systems, the amount of data processed by IDS must be reduced to achieve low detection latency. In this paper, we investigate the problem of feature selection in medical cyber physical systems. Our initial results demonstrate the laplacian scoring techniques are successful in optimal feature selection with reduced memory consumption.