Franz Papst will present an online talk within the seminar “Analysis of Complex Systems” on June 26, 2021, 3PM-4PM (CET) via Zoom.
If you would like to attend, please email email@example.com
Title: On Location Privacy in IoT Sensor Data
Data sharing is crucial for building large data sets in different fields of application. Data privacy is a legitimate concern of subjects contributing to the data. Even when all unique identifiers like names or identification numbers are removed from a dataset, the dataset is not anonymized but rather pseudonymization. It is still possible to identify individuals in a given pseudonymized dataset, when linking pseudonymized data with publically available data. This is especially true for time series sensor data, which consist of a sequence of data points rather than just a single data record.
In this talk, I am going to show how to localize the origin of sensor data by combining it with publically available weather data. With our method we are able to locate cows with an average localization accuracy of 28.2 km over the area of Austria with nothing more than a trace of sensor data measured in their rumen. For sensor readings from O3 data and solar data, we are even able to lower the average error to 25.9 km and 7.4 km respectively.
I will illustrate how to perturb the data to avoid this kind of privacy leakage. I will also illustrate how to tune the privacy-utility trade-off.