This task will start with the study of the state of the art event models and analyse them with respect to the requirements identified.
This task will then be decomposed in the following three activities:

1. Service Social Modeling: the aim of this activity is to propose a new social model that allows representing and reasoning on the various social links between services. This model will be used to represent social preferences expressed in the subscription to service events. Social links between services can be of different types: explicit links (e.g., declared trust between two services that interacted in the context of a service composition), implicit links (e.g., the dependency between two services that are often used in conjunction with each other). For the creation of the social service modelling, we will investigate the extension of the Friend Of A Friend (FOAF)4 ontology that have been developed for modelling online social networks.
2. Social-aware event subscription: the aim of this activity is investigate an extension of existing event models (e.g., WS Notification) in order to support the specification of :
   • Event Semantics : designates formal information about the content of the event
   • Event Temporal Properties : designates temporal information about the event
   • Event Spatial Properties : designates contextual information within a specific location or area
   • Event Social Service Preferences : designates contextual information related to social preferences between services
3. Social-aware event matching: the aim of this activity is to propose event matching algorithms that consider social service preferences in addition to semantic, temporal and spatial event attributes.

Architecture and Functionality

The Social Filter operates on a social network of services (or nodes) to compute the strength of the relationships between them. The social filter provides an interface to the Relationship Strength Engine component. The Event Cloud calls the social filter and submits a source node and a list of target nodes. The relationship strength engine uses a trust inference algorithm to compute the strength of the relationships between the source node and each target node.

The first possible case is that the source node has a direct social relationship with all target nodes. In this case the social filter computes relationship strength as a weighted sum of normalized characteristics of the relationship between the two nodes. These characteristics include the time of last interaction, interaction count, declared trust, number of mutual nodes, etc.

The second possible case is that no direct social relationship exists between the source node and one or more of the target nodes. In this case the social filter computes relationship strength as the maximum flow of declared trust from the source node to a target node in the social network.

The social filter returns the computed strength of the relationships back to the event cloud. The event cloud may use a relationship strength threshold or a ranking of the destination nodes to select the most trustworthy nodes. Consequently the event cloud can accept to process the events that the source node received from target nodes with which it has high relationship strength whereas it can discard the ones received from target nodes with low relationship strength.

The Social Editor allows a user to view and update the social network.

Demo