Applying Reversible Privacy to Streaming Data
Koivunen, Lauri (2019-08-12)
Applying Reversible Privacy to Streaming Data
(12.08.2019)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
suljettu
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2019082625562
https://urn.fi/URN:NBN:fi-fe2019082625562
Tiivistelmä
Even though the world is full of data with great potential to improve our living, a lot of it is out of public reach and also not available for research purposes due to security and privacy concerns. A special case of datasets, streaming data, can be a potential goldmine for research purposes, but its challenging privacy aspects are often neglected. This thesis applies privacy enhancing transformations to a large stream of structured data to allow a degree of privacy for the subjects depicted in the data while still making it possible to use the data for research purposes.
The privacy enhancing transformations are applied in such a way that any discoveries of anomalous data could be traced back to the original records, even after applying transformations. In medical data this property is used to the advantage of the study subjects so that they could be set to benefit from any found results if desired. In the scope of this thesis, the property of reversibility is used to trace back identified malicious malware activity to clean possible infections.
The results of this thesis include a streaming data anonymizing system tailored for our research input data. The various privacy enhancing transformations applicable on streaming data are outlined and analyzed in relation to the input data. Tradeoffs between anonymity, data quality and size are also discussed for each transformation. A framework for validating the anonymity of these transformations is discussed. An analysis of shortcomings of the case study implementation is also provided.
Streaming data privacy can be enhanced from raw data, but not all of the transformations that an offline database could afford can be substituted with stream-aware alternatives. Being able to guarantee both perfect anonymity and a dataset of equal quality to non-anonymized version is most likely not possible.
The privacy enhancing transformations are applied in such a way that any discoveries of anomalous data could be traced back to the original records, even after applying transformations. In medical data this property is used to the advantage of the study subjects so that they could be set to benefit from any found results if desired. In the scope of this thesis, the property of reversibility is used to trace back identified malicious malware activity to clean possible infections.
The results of this thesis include a streaming data anonymizing system tailored for our research input data. The various privacy enhancing transformations applicable on streaming data are outlined and analyzed in relation to the input data. Tradeoffs between anonymity, data quality and size are also discussed for each transformation. A framework for validating the anonymity of these transformations is discussed. An analysis of shortcomings of the case study implementation is also provided.
Streaming data privacy can be enhanced from raw data, but not all of the transformations that an offline database could afford can be substituted with stream-aware alternatives. Being able to guarantee both perfect anonymity and a dataset of equal quality to non-anonymized version is most likely not possible.