Communications of the ACM
Social Network Analysis Privacy Tackled
Sofya Raskhodnikova suggests "differential privacy" is needed to maximize accuracy of analysis while preventing identification of individual records.
Credit: iStock Photo SbytovaMN
Pennsylvania State University (PSU) researchers are researching ways to maintain privacy on social networks.
PSU professor Sofya Raskhodnikova says the goal is to be able to release information without making personal or sensitive data available and still be accurate.
With multiple public databases available, data can easily be correlated between databases to assemble pieces of deleted data and recover the identifying information. Differential privacy, which restricts the types of analyses that can be performed to those for which the presence or absence of one person is insignificant, is needed to maximize the accuracy of analysis while preventing identification of individual records. Differential privacy guarantees an analysis performed on two databases that differ in only one record will return nearly the same result.
Raskhodnikova says, "one approach for achieving differential privacy is adding a small amount of noise to the actual statistics before publishing them," but the problem is determining how much noise and how to execute it so the accuracy of results is retained.
The idea of differential privacy could be especially important to the protection of graph data. The researchers found differentially private methods for releasing many graph statistics.
The degree distribution of a social network specifies how many friends each member has, but some information is inherently too sensitive to be released with differential privacy.
From Penn State News
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found