How to Securely Perform Computations on Secret-shared Data
Name
Dan Bogdanov
Abstract
Databases containing personal, medical or financial information about an individ-
ual are usually classified as sensitive. Often the identity of the person is somehow
stored in the database, whether by name, personal code or a combination of at-
tributes. In many countries it is illegal to process such data without a special
license from the authorities. Such protection is needed to preserve the privacy of
individuals and prevent abuse of the data.
This level of protection is a problem for research organisations, that cannot
learn global properties or trends from collected data. It also prevents government
organisations from providing accurate demographics reports and managing medical registries about the population. Data analysis restriction is not the only problem for these organisations but a solution is nevertheless expected.
In this thesis we address a simplified version of the problem. Assume that we
have asked p people q sensitive questions. By collecting the answers we obtain a
matrix D with p rows and q columns denoted that represents our data. Our goal
is to devise a method for computing aggregate statistics from this matrix without
compromising the privacy of a single person, that is, revealing values in matrix
D.
Graduation Thesis language
English
Graduation Thesis type
Master of Science in Engineering (4+2) Computer Science*
Supervisor(s)
Jan Villemson, Sven Laur
Defence year
2007