Trust and reputation systems are widely employed in WSNs to help decision making processes by assessing trustworthiness of sensors as well as the reliability of the reported data. Iterative filtering (IF) algorithms hold great promise for such a purpose; they simultaneously estimate the aggregate value of the readings and assess the trustworthiness of the nodes. Such algorithms, however, operate by batch processing over a widow of data reported by the nodes, which represents a difficulty in applications involving streaming data. In this paper, we propose STRIF (Streaming IF) which extends IF algorithms to data streaming by leveraging a novel method for updating the sensors' variances. We compare the performance of STRIF algorithm to several batch processing IF algorithms through extensive experiments across a wide variety of configurations over both real-world and synthetic datasets. Our experimental results demonstrate that STRIF can process data streams much more efficiently than the batch algorithms while keeping the accuracy of the data aggregation close to that of the batch IF algorithm.
展开▼