Secure data fusion is an important application of wireless sensor network. Previous works are implemented by authentication among neighboring sensors. Considering that in many cases, data fusion event is both spatially and temporally correlated, this paper proposes a distributed and localized secure data fusion algorithm based on behavior trust. The security and reliability of the algorithm are implemented by statistical hypothesis test for matching the reading sequence of sensors and statistical characters of the data fusion event. The simulation results show the validity of the algorithm. Keywords wireless sensor network; data fusion; behavior trust; security
INTRODUCTIONAn important character of wireless sensor network is that a set of sensor nodes with limited function complete a great perceptive assignment concurrently. Due to the sensor network nodes have limited battery energy, disposal ability, storage capability and communication bandwidth, they are influenced by environment greatly. We should use the nodes' ability of local calculation and storage ability to dispose data, get rid of redundant information, reduce transmission quantity and dispose information fusion. [1][2][3] In wireless sensor networks, the functions of data fusion are mostly showed on saving energy, improving data collection efficiency, enhancing data accuracy and getting synthesis information, etc. Data package fusion mostly combines with routing protocol on network layer. The typical algorithm includes Data funneling [4], AIDA[5] and so on. At the same time, on application layer, we can use search technique of distributed database to filter collected data step by step. The typical algorithm includes TAG [6], TINA[7] , Medians and beyond[8], etc. Besides, there is a model driving fusion that uses some model to describe and fuse data. The typical algorithm includes Wavelet neural -networks [9], Kalman filtering[10], Probabilistic model [11]and so on.Be different from traditional multi-sensor data fusion, sensor networks' character decides the complexity in researching data fusion. An important thing to achieve data fusion in wireless sensor network is to make sure security and credibility during data fusion. Presently, previous data fusion algorithms in sensor networks mainly use the methods based on encryption and authentication to achieve security, they can not dispose data fusion's insecure problems resulted from false behaviors as it said [12]. In recent years, networks' trust research has become a hotspot [13]. Networks' trust technology is a new secure method which adds behavior trust based on the previous networks' secure technology, intensifies dynamic disposal on network states, provides strategy basic for implementing intelligent self-adapted networks' security and service quality control.