When you are entering a dark building at night, the sudden lighting at the door--the sensors at work--may startle you. On a highway, you can speed through a tollgate without stopping because an electronic toll collection (ETC) system senses your car passing and collects the fee automatically. From the above examples, you can see that sensors and their various applications are everywhere. If sensors are equipped with wireless transmission and reception capabilities, they can form a wireless sensor network (WSN).

In a WSN, sensors monitor a continuously changing situation--for example, thermal sensors monitor temperature changes and manometers monitor pressure changes--and variations in the monitored data are usually normal. However, changes also occur when there are intruders in the WSN or faults occur. This paper focuses on how to determine if a change in the sensed data is normal or abnormal.

In the abstract, the authors describe their work: “We develop a series of Markov models, including tree-indexed Markov chains which can model its spatial structure. For each model, an anomaly-free probability law is estimated from past traces.” Although the authors run simulations for validation, the simulation parameters and environments are not clearly stated.

Overall, the paper is more about mathematical Markov model analysis than about sensor networks, which just happen to be an area where the Markov model can be applied. The paper could use more examples and real numerical data or traces that illustrate how their model can be applied.