452 PART VI Event Monitoring and Situation Awareness In particular, Section 2 of this chapter dis- cusses the state-of-the-art on sensor event anal- ysis for health monitoring. The authors point out that most existing work only supports the recognition of pre-defined interesting patterns and requires significant amounts of training data. These limitations, however, may significantly limit the applicability of existing techniques to real-world health monitoring systems. A detailed review of the existing work and the associated problems is then made available in Section 3. In Section 4, the authors discuss two key prim- itives for discovering patterns of interest: (1) the discovery of frequent discontinuous sequences ­ i.e., sequences of events which may not happen right next to each other ­ from the input stream of activities, and (2) clustering sequences into groups of activities based on the results from the first step. Then, in Sections 5 and 6, the authors discuss the discovery of interesting activ- space and computational resources to process a continuous stream of large amounts of sensory data coming in at a high speed and to achieve provable error bounds. This chapter reviews the existing algorithms that offer bounded running time and small memory footprints ­ i.e., algo- rithms that are feasible to be used in real-time distributed monitoring systems. After an overview of the challenges for sys- tem monitoring in cyber­physical infrastructures, Section 2 of this chapter provides a model of distributed system monitoring for cyber­physical infrastructures. Also in this section, the authors define the data collected by monitoring sensors as a data stream and outline the objective of data stream algorithms in terms of the processing to be performed over a sliding window of the incoming stream, as well as the space and time overhead constraints. Section 3 of this chapter reviews in detail