Selective Reprogramming of Mobile Sensor Networks through Social Community Detection

We target application domains where the behavior of animals or humans is monitored using wireless sensor network (WSN) devices. The code on these devices is updated frequently, as scientists acquire in-field data and refine their hypotheses. Wireless reprogramming is therefore fundamental to avoid the (expensive) re-collection of the devices. Moreover, the code carried by the monitored individuals often depends on their characteristics, e.g., the behavior or preferred habitat. We propose a selective reprogramming approach that simplifies and automates the process of delivering a code update to a target subset of nodes. Target selection is expressed through constraints injected in the WSN, triggering automatic dissemination of code updates whenever verified. Update dissemination relies on a novel protocol exploiting the social behavior of the monitored in- dividuals. We evaluate our approach through simulation, using real-world animal and human traces. The results shows that our protocol is able to capture the social network structure in a way comparable to existing offline algorithms with global knowledge while allowing runtime adaptation to community structure changes, and that existing dissemination approaches based on gossip generate up to three times more network overhead than our socially-aware dissemination.