Toward Evaluation of Deployment Architecture of ML-Based Cyber-Physical Systems

In this position paper, we explore a way to assess deployment architectures for intelligent Cyber-Physical Systems (iCPS). Our long-term goal is to establish a method that allows a software architect of iCPS to choose an appropriate deployment architecture, based on requirements on quality, cost and security. As the first step towards the goal, this paper first presents an abstract model of iCPS to capture the possible deployment configurations. We then propose a method that assesses cost-related attributes for the configurations. Finally, we investigate security threats within different configurations, to help the engineer to achieve the configuration choice.