Methods for assessing trustworthiness on AIoT systems: an SLR

Sammanfattning

The integration of artificial intelligence (AI) with the Internet of Things (IoT) has given rise to the Artificial Intelligence of Things (AIoT), enabling intelligent, autonomous, and interconnected systems. However, the trustworthiness of AIoT systems remains a critical challenge, as they must operate reliably under dynamic conditions while ensuring security, privacy, and performance. This research addresses the question: How is trustworthiness assessed in AIoT systems, and what metrics and characteristics are most frequently applied? A systematic literature review (SLR) of 67 peer-reviewed papers was conducted, extracting and categorizing assessment methods, metrics, and key characteristics. The analysis reveals that quantitative simulation-based assessment is the most prevalent method, followed by blockchain-based trust modelling and conceptual/hybrid frameworks. Metrics such as accuracy, security level, privacy level, and attack detection rate are dominant, while decentralization, performance modelling, and resource efficiency are key recurring characteristics. These findings contribute to a structured understanding of trustworthiness evaluation in AIoT and inform future research and standardization efforts.