Efficient Top-k Monitoring of Abnormality in Sensor Networks

There exist challenges in continuous monitoring application, mainly due to the limited battery power of sensor nodes all the time. Power efficiency is the key. A power efficient top-k monitoring framework called PECTMA is proposed which is a three level architecture including four novel algorithms, CRVMR, ESR, top-k-sort and BRCR. The basic idea is to install two level filters. One is at each sample sensor node, called CRVMR, to save power battery by suppressing unnecessary data transmittance with a regression function. Another is at each cluster, called ESR, to further reduce the communication cost by eliminating the spatial redundancy. The performance of PECTMA is evaluated using synthetic data sets which is get by referencing the real data sets. Experiment results show that PECTMA is substantially outperforms the existing filter-based TA and the TAG-based approaches for continuous top-k abnormal monitoring when the sample reading are inherently fluctuates, especially with constant periodicity.