An architecture for socially mobile collaborative sensing and its implementation

Abstract

We introduce the concept of "socially mobile collaborative sensing" (SMOCS), a collaborative process in which devices owned by different users gather and share contextual data and derive inferences. People are increasingly gathering contextual data (e.g. location [38]) for their own use and are beginning to share it through systems such as Twitter and Google Latitude. These systems are domain specific (they are designed for a single data type) and require all users to gather their context themselves; users must install software on a device capable of gathering rich contextual data (e.g. a smartphone). The SMOCS architecture enables users with less-capable devices to offload contextual sensing and reporting to physically proximate more-capable devices; in this way, SMOCS defines a truly viral architecture. Using collaboration among devices that are found in the neighborhood of a mobile user to gather contextual data is now practical. The density of reachable devices in any particular person's neighborhood is growing with time. SMOCS is motivated by the need to structure such collaborations to support viral growth and evolution of applications that exploit such sensing. In this thesis we define SMOCS and present an architecture and an implementation, called ContInt (Context Interleaving). The ContInt implementation is composed of two components: Ego, a distributed social network in which users maintain personally-owned "agents", and the ContInt plugin for Ego.

(cont.) The proposed SMOCS architecture enables the collection and distribution of contextual data and provides extensible interfaces to allow inference-deriving "plugins." We evaluate ContInt in terms of a) scalability and performance, b) architectural extensibility and c) argue that its privacy model enables users to control their data. We conclude by proposing future work and our expectations of how SMOCS might evolve.