Sarcouncil Journal of Engineering and Computer Sciences

Abstract: The low-power IoT sensor streams in environmental networks are characterized by complex physical processes, various measuring modalities and rich temporal dynamism. Sparse sensors would like to send low-rate volume records subject to time and space constrains. Behavioral anomalies, with anomalous (i.e., atypical) increments to an otherwise pristine low-activity pattern, emerge as a heuristic compact description of these data streams. The visible granularity in atomic behavior indicates that an irregularity-driven representation might support lightweight cross-dataset theoretical examination of anomalous behaviors. The environmental communities have been identified from three datasets: Intel Lab Data, SensorScope and Gas Sensor Array Drift -- which capturing similar streams. A behavioral anomaly characterization for low power IoT traffic is presented in conceptual terms. Sensor streams possess features associated with their environmental nature which anomalies are based on, therefore causing a general observational interest. Anomalies are handled qualitatively through the framework of theoretical modeling that allows to adopt a common language for irregularity signatures rather than to include dataset-specific empirical numbers.