I know what you did in the ICT-WG
Smart home solutions produce a lot of data with their sensors and so is the case for the various smart home systems we have in our shared apartment. Fortunately, most of the systems do not store the data and hence do not keep track of the proceedings. Thinking about this, we started to wonder what can be extracted if the data would have been stored and decided to regularly upload the state of all sensors to the cloud.
Collected Data and Method
Only the common rooms, such as the living room or the kitchen, are considered for the data collection, since including our personal room might be an unwanted insight in our privacy. An independent openHAB system has been set up for the sole purpose of getting the sensor states and uploading it to the my.openHAB cloud. At the moment, the dataset includes motion, brightness, temperature, humidity, door-switches and lights. In a next step, it is planned to also include electronic devices, such as TV or the Sonos sound system, domestic appliances and also weather in order to correlate some events to the conditions outside. It is also planned to upload the data to the Swisscom IoT-cloud and to our NAS server to have redundancy.
Goal of this Project
With this project we would like answer the following set of questions:
- What can be read out of the data?
- Is it possible to see how many people were present at home and even further to see what they were most likely doing?
- Is it even possible to characterize the people living in the apartment?
- If someone gets access to our data, how much of our privacy are we disclosing with the datasets and can this be used in a malicious way?
- Can the data be used to design a self learning smart home solution?
- What is the percentage of regularities in the dataset?
- Are these regularities only specific to a shared apartment with three male inhabitants or are the applicable to any kind of apartment?
- How much time would a self learning system need to correctly identify these regularities?
We are aware that a few months of data is not enough to fully answer these questions, but we hope that we can see certain trends that can lead to a partial answer.