Towards Cost-Benefit-Aware Adaptive Monitoring for Cyber-Physical Systems

Abstract

Cyber-Physical Systems (CPS) are becoming ubiquitous in many different domains, for example, in the form of Unmanned Aerial Vehicles (UAVs), (semi-)autonomous systems, or robotic applications. Given that CPS frequently operate in a safety-critical context, and interact and collaborate with humans, ensuring that these systems behave as intended and adhere to their specified security and safety requirements at runtime is essential. Providing automated support for monitoring is a fundamental part of collecting information about the system, and facilitating subsequent analysis and reasoning. However, while advances have been made, particularly in self-adaptation and self-management, the monitoring aspect is often neglected, resulting in suboptimal data collection that does not consider associated monitoring costs, changing environments, or varying benefits of the collected data. Such benefits range from accountability in the aftermath of a security incident, up to proactive defense against risks if connected to alarming. In this paper, we outline our initial concept for costbenefit-aware adaptive runtime monitoring, for (but not limited to) safety and security requirements. As part of this work, we identify relevant monitoring aspects and create a Cost-Benefit-Aware Adaptive Model (CBAAM), conceptualizing the costs and benefits of adaptive monitoring. This is also especially relevant for security, saving resources, e.g., budget and computational. We further present an architecture for defining and executing these adaptations and discuss our research roadmap and next steps.