Hagedoorn, Matthijs (2026) Inferring Sensitive Personal Information from Google Location History Using Data Visualization Techniques. Bachelor thesis, Data Science and Society (DSS).
|
PDF
DSSs5743826MHagedoorn.pdf Download (6MB) | Preview |
Abstract
Abstract This study investigates the extent to which sensitive personal information can be inferred from an individual's Google Location History (GLH) using data visualisation techniques. Unlike prior work on location privacy, which largely relies on complex models applied to large or anonymised datasets, this study operates on a single personal export file and asks what becomes readable through visual inspection alone. The GLH export is itself the product of substantial platform-side processing: raw signals are converted into semantically labelled visit records, confidence scores, and place identifiers before the file reaches the user. The contribution of this study lies in showing that, at that point, no additional modelling is required to recover sensitive information. Using three years of personal location data, two visualisation tools were developed: an interactive kernel density heatmap with hourly and weekday/weekend disaggregation, and a static three-panel temporal patterns figure. Together these were sufficient to identify home address, occupational location, habitual commute route, sleep schedule, weekly routine, regular social and leisure venues, and a recurring out-of-city overnight destination. A peer classification exercise in which five participants independently classified the top 15 most-visited locations produced majority agreement on all 15, suggesting the visualisations carry inferential value independently of prior familiarity with the researcher. Three findings stand out. First, the home location is identifiable from GPS density and timing alone, without any semantic labelling. Second, the high proportion of UNKNOWN labels in Google's own classification, covering 75.6% of visit records, does not prevent inference and does not function as a privacy protection. Third, the entire analysis relied only on standard open-source Python libraries and a freely available export file, placing this capability within reach of any motivated person with basic programming skills. These findings suggest that privacy risk from personal location data exports extends well beyond sophisticated actors to anyone with access to a Google Timeline file and a map.
| Item Type: | Thesis (Bachelor) |
|---|---|
| Name supervisor: | Schauble, J.K. |
| Date Deposited: | 10 Jun 2026 11:42 |
| Last Modified: | 10 Jun 2026 11:42 |
| URI: | https://campus-fryslan.studenttheses.ub.rug.nl/id/eprint/810 |
Actions (login required)
 |
View Item |