Privacy-preservation and calibration of smart sensors

New and affordable sensors (e.g., Arduino Nano 33 BLE Sense, Sensebox) have greatly increased the potential to collect finer-grained environmental information, especially on topics impacting on human health, e.g., noise and air pollution. Using the sensors raises at least two important issues: First, the sensor readings might not be of the same consistent accuracy as that of more expensive sensors that are placed and maintained according to strict protocols. Second, continuously recording and storing centrally various environmental data from a dense network impacts on privacy.

This proposed topic investigates those issues while keeping the geospatial aspect in focus. The aim is to design a geospatial sensor network and implement the prototype of a single sensor that is continuously calibrated and does not intrude on privacy. A suggested solution is a form of edge computing that utilizes a pre-trained machine-learning model.

Depending on the MSc program and ECTS load of the thesis module, the focus can be on on both or just one of the two issues. In any case, required skills include programming (learning C++ for programming the sensors is doable within the time frame of a thesis research only if programming skills in , e.g., Python already exist) and quantitative analysis skills (for training and interpreting the machine-learning model using Python).