Cohort-level protection and individualized inference in artificial intelligence-based monitoring applications

Abstract

In several modern application domains, artificial intelligence (AI)-driven automated tools are used for monitoring the functionality, efficacy and performance of a real-world system. Examples include healthcare monitoring for cancer, neurodegenerative diseases and so on, financial monitoring for fraud or security breaches or insider activities, cybersecurity applications, environmental and ecological monitoring and protection, industrial production and control, traffic and urban utilities protection and monitoring and numerous other applications. In many of these cases, a large cohort of similar units may be monitored; however, automated decisions are required at an individual or unit level. Our studies show that an entropy-based metric used in a process control framework leads to early and accurate detection of breach of normal pattern, and can result in secure and reliable use of AI-driven protection and monitoring.*The full version of the paper is in preparation and will be made available online soon.