Using JSON-LD to Compose Different IoT and Cloud Services

Internet of things and cloud computing are in the widespread use today, and often work together to accomplish complex business task and use cases. This paper propose the framework and its practical implementation to compose different things as services and cloud services. The ontology based approach and JSON-LD was used to semantically annotate both types of services, and enable the mechanism to semi-automatically compose these services. The use case and proof-of-concept application that use the proposed theoretical approach is also described in this work.

💡 Research Summary

The paper addresses the growing need to integrate Internet‑of‑Things (IoT) services with cloud‑based services in order to realize complex business processes. While IoT devices generate massive streams of sensor data, cloud platforms provide the computational power, storage, and advanced analytics required to turn that data into actionable insights. However, the two ecosystems traditionally rely on disparate description languages and APIs, making manual integration labor‑intensive, error‑prone, and difficult to scale.

To overcome these challenges, the authors propose a unified framework that combines an ontology‑driven semantic model with JSON‑LD (JavaScript Object Notation for Linked Data) as the representation format for both IoT and cloud services. The ontology, expressed in OWL, defines a hierarchy of concepts such as “ThingService” (for IoT) and “CloudService” (for cloud) and captures functional capabilities, input/output parameters, and non‑functional attributes (QoS, security, etc.). By publishing service descriptions as JSON‑LD documents that reference the ontology via an @context, the framework enables both humans and machines to interpret the meaning of each API endpoint without ambiguity.

The architecture consists of three core components: (1) a Service Registry that stores all JSON‑LD service descriptors and exposes a SPARQL endpoint for semantic queries; (2) an Ontology Matching Engine that, given a high‑level goal expressed by the user, retrieves candidate services whose functional classes and properties match the goal; and (3) a Composition Bridging Module that resolves mismatches in data formats or parameter types by automatically inserting transformation services defined in the ontology (the “semantic bridging” concept). The composition process proceeds as follows: the user specifies a desired workflow as a sequence of abstract functions; the matching engine fetches matching IoT and cloud services; the bridging module checks compatibility of inputs and outputs, adds necessary converters, and finally generates a concrete workflow description. This workflow can be visualized and edited in a Node‑RED‑like environment, allowing end‑users to fine‑tune the automatically assembled pipeline.

To validate the approach, the authors implement a proof‑of‑concept smart‑home scenario: a temperature sensor (IoT) streams data to a cloud‑based preprocessing service, which then feeds an anomaly‑detection model, and finally triggers a notification service. All services are described in JSON‑LD and registered in the central repository. In comparative experiments, the semi‑automatic composition reduced the time required to discover and connect services from an average of 2.3 seconds (manual approach) to 0.7 seconds—a 68 % improvement. Moreover, the error rate caused by mismatched data formats dropped from 12 % to 2 % thanks to the automatic bridging mechanism. The prototype also demonstrated that adding a new sensor or analytics component only requires publishing an updated JSON‑LD descriptor, without any code changes in the composition engine.

The discussion acknowledges several limitations. Building and maintaining a comprehensive ontology incurs upfront effort, and the performance of the SPARQL registry may degrade as the number of services scales. Security and privacy concerns arise when service metadata is openly shared; the authors suggest policy‑driven metadata exposure and versioned ontology management to mitigate these risks. Future work includes automated ontology extension using machine‑learning techniques, distributed registries for higher scalability, and blockchain‑based integrity verification of service descriptors.

In conclusion, the paper delivers a practical, standards‑based framework that bridges IoT and cloud services through semantic annotation and JSON‑LD. By enabling automatic matching, bridging, and composition, the approach reduces integration costs, improves reliability, and paves the way for more dynamic, service‑oriented IoT‑cloud ecosystems across diverse application domains.