This thesis explores the use of air quality sensors to infer environmental context, such as distinguishing between indoor and outdoor spaces or identifying different types of surroundings. A custom-built multi-sensor platform was developed to measure CO2 , humidity, temperature, pressure, particulate matter, and light. After applying robust preprocessing techniques, the data was segmented and evaluated using both unsupervised clustering and supervised classification.

K-Means clustering and PCA helped assess the natural separability of the data, while a Random Forest model was used to classify our data. Feature importance analysis showed that the light, CO2 and fine particle data were among the most informative signals. The findings highlight the potential of gas sensor data for location inference, but also raise privacy concerns, particularly in light of increasing deployment of smart devices in residential spaces. This underscores the importance of considering ethical implications alongside technical feasibility.