HTTP Traffic Datasets for Research in Service-Oriented Computing

We present three HTTP datasets for experimenting on various aspects of service-oriented computing.

The datasets were generated by creating random traffic targeting the services offered by Google Tasks, Slack, and Twitter. In order to form transactions, various operations to create, read, update and delete (CRUD) service-specific resources were created and the respective responses were recorded, simulating service interactions by users through applications. The resources the operations interacted with are lists (Google Tasks), messages (Slack) and tweets (Twitter). 

The input generation used fuzzing techniques. In particular, Apache JMeter was used as it has the functionality to fuzz RESTful services (randomly generate various types of API calls by providing different inputs) and recording interactions in a suitable textual format for further processing. The fuzzing was guided by a light-weight semantic service model provided as Swagger (recently renamed as OpenAPI) spec. 

The datasets reflect the richness of modern Web APIs for experiments and are in the XML format.