With extremely limited resources, it is hard to protect sensor networks well with conventional security mechanisms. We study a class of passive fingerprinting techniques and propose an innovative
based approach to improve the security of sensor network. For each sensor, we create a quality profile
of profiling its normal/standard sensing behaviour. After deployment, new sensor readings are verified using this
. If significant deviation is found, we either regard the readings as an abnormal behaviour or declare the sensor to be a fake sensor. The methods can be used as an assistant sensor authentication mechanism, but with a potential drawback. Furthermore, we also demonstrate a secure data fusion protocol, applying the proposed methods together with conventional security mechanisms. Through security analysis, we point out several countermeasures that can explicitly or implicitly defend against these attacks.