Analyzing Breaking Changes in IoT Systems: A Taxonomy and Empirical Study on System Stability and Longevity

Evolving application programming interfaces (APIs) of Internet of Things (IoT) devices cause system failures and challenge the longevity and reliability of IoT systems. To mitigate these issues, this paper aims to categorize IoT API changes and assess their impact on deployed systems. We introduce a detailed taxonomy that classifies API changes into 28 characteristics within five key dimensions: Data Payload Modifications, Communication Protocol Modifications, API Endpoint Modifications, Security Modifications, and Parameter Modifications. We apply this taxonomy by annotating and analyzing a comprehensive dataset developed through an empirical study of change reports and integration documentation from long-running open-source IoT integration platforms. We analyze the frequency and characteristics of evolving IoT APIs and the effects of changes within IoT device integration scenarios. Results reveal that 74.21% of all API changes cause backward incompatibility, with Data Payload and API Endpoint Modifications being the most prevalent types, and that cloud-based platform integrations are most vulnerable to changes that threaten system longevity. This work contributes to IoT system design by providing a comprehensive taxonomy of IoT API changes, facilitating the development of more robust and adaptable IoT architectures that can withstand long-term operational changes and enhance system reliability.