Analyzing the behavior of sensors is becoming one of the key challenges due to their increasing use for decision making in IoT systems. The paper proposes an approach for a formal specification and analysis of such behavior starting from existing sensor traces. A model that embodies the sensor measurements over time in the form of stochastic automata is built, then temporal properties are fed to Statistical Model Checker to simulate the learned model and to perform analysis. LTL properties are employed to predict sensors’ readings in time and to check the conformity of sensed data with the sensor traces in order to detect any abnormal behavior. We also use LTL properties to analyze the collective behavior of a set of sensors and build a formal model that checks the conformity of a combination of sensors’ readings in time.

由于传感器越来越多地用于物联网系统中的决策，因此分析传感器的行为正成为关键挑战之一。本文提出了一种从现有传感器轨迹开始对此类行为进行正式规范和分析的方法。构建一个模型，该模型以随机自动机的形式体现随时间的传感器测量，然后将时间属性馈送到统计模型检查器以模拟学习模型并执行分析。LTL 特性用于及时预测传感器的读数并检查感测数据与传感器迹线的一致性，以检测任何异常行为。我们还使用 LTL 属性来分析一组传感器的集体行为，并构建一个正式模型来及时检查传感器读数组合的一致性。