Evaluating Practical Privacy Attacks for Building Data Anonymized by Standard Methods

Jens Hjort Schwee*, Fisayo Caleb Sangogboye, Mikkel Baun Kjærgaard

*Corresponding author for this work

Research output: Contribution to conference without publisher/journalPaperResearchpeer-review

215 Downloads (Pure)


The availability of inexpensive IoT sensors enables the collection of a wide range of data about buildings and their use by occupants. The data originating from these sensors are in many cases privacy-invasive. Therefore, methods for improving privacy protection for such data is needed. In this paper, we explore if suppression, k-anonymity or a combination of them, provides sufficient privacy protection on a published dataset. The results show that the specific dataset only was partly protected in the case of the combination of the two privacy protection methods. The attack vectors used to break the protection includes the sensor deployments, and the physically limitations of the sensors. This indicates, the need to consider how to protect the metadata of sensor deployments as well as the sensor streams. This additional metadata protection can serve as requirements, for anonymization methods for time-series data. However, even with good privacy protection, the building dataset might still be vulnerable to attacks since building data contains repeatable patterns which are susceptible to a range of attacks.
Original languageEnglish
Publication date15. Apr 2019
Number of pages4
Publication statusPublished - 15. Apr 2019
EventInternational Workshop on Security and Privacy for the Internet-of-Things - Montreal, Canada
Duration: 15. Apr 201915. Apr 2019
Conference number: 2


WorkshopInternational Workshop on Security and Privacy for the Internet-of-Things
Internet address


  • Data privacy
  • Pseudonymity
  • Data anonymization and sanitization
  • k-anonymity
  • Privacy-protecting data publishing
  • Linkage attacks


Dive into the research topics of 'Evaluating Practical Privacy Attacks for Building Data Anonymized by Standard Methods'. Together they form a unique fingerprint.

Cite this