Log Analyzer for IoT Applications

Internet of Things applications touch numerous facets of our lives with a great potential for exponential growth. This implies the rapid growth of the amount of newly generated and collected data. Consequently, the demand to have immediate insights into the data and to turn this immense amount of generated data into valuable information increases simultaneously. In the world of IoT networking, most devices are connected to the internet and realize machine to machine (M2M) communication amongst themselves. A sensor or machine sends recorded data to the application software that processes the data for further usage. In most cases, all the steps or events occurring in an IoT system are recorded and gathered in log files.

One possible approach to turn data into actionable insights and these insights into value is to analyze historical data produced by IoT applications, such as log files. This thesis presents a Log Analyzer for IoT applications with the use of stochastic methods. The log file entries of an IoT application can be considered as a collection of process step variations, so-called sub-sequences. This thesis presents an approach to find remarkable deviations between the analyzed sub-sequences and compare sub-sequences according to different similarity measures. In addition, an approach to visualize a graphical model of the underlying dynamic process, which represents a well-defined probabilistic model showing sets and exact sequences of states, is presented.