Status : Verified
| Personal Name | Mendoza, Marie Jo-anne M. |
|---|---|
| Resource Title | Varying Privacy Budgets for Context-Aware Local Differential Privacy |
| Date Issued | June 2018 |
| Abstract | A user’s participation in a study leads to his/her personal and possibly sensitive data to be stored in a statistical database, where data analysts can perform calculations and then extract useful information. Privacy is guaranteed through generic, one-size-fits-all privacy policies defined by service providers, which could still leave the data of the end-users vulnerable. This problem is solved by -differential privacy, wherein noise (as a function of the constant privacy parameter ) is added to the aggregated data to protect individual users and provide them an avenue to deny their participation in the study. However, not all pieces of data have the same weight, and users may also have differing definitions of privacy and how much risk they are willing to take in case their data is exposed. This study looks at -differential privacy when applied locally, or on the side of the users, so that they could have full control over how much of their real data they are actually giving up, before it even reaches the ones collecting the data. By varying the distribution of cautious users (who require more privacy and a lower) to those more tolerant to risk (higher ), we see which local differential privacy mechanisms were as effective as the centralized differential privacy mechanisms when applied to a particular type of variable, and which are not. |
| Degree Course | Master of Science in Computer Science |
| Language | English |
| Keyword | privacy; differential privacy; aggregate data; laplace mechanism; exponential mechanism |
| Material Type | Thesis/Dissertation |
Preliminary Pages
5.70 Mb
Category : F - Regular work, i.e., it has no patentable invention or creation, the author does not wish for personal publication, there is no confidential information.
Access Permission : Open Access