CS Researchers Win Best Paper Award at ACM CODASPY 2016

03-30-2016
Writer(s): Staff Reports

Researchers from the Purdue Computer Science Department, Institute for Infocomm Research and Samsung Research America won the Best Paper Award at the 6th ACM Conference on Data and Application Security and Privacy conference, recently held in New Orleans, LA.

The paper, "Differentially Private k-Means Clustering" is co-authored by Ph.D. candidate Dong Su and his advisor, Professor Ninghui Li, in collaboration with Professor Elisa Bertino, Dr. Jianneng Cao from Institute for Infocomm Research, and Dr. Hongxia Jin from Samsung Research America.  

The paper describes a new approach for releasing data for k-means clustering with high accuracy while preserving the privacy of individuals whose data are used.  k-means clustering aims to partition a set of observations (data points) into k clusters in which each observation belongs to the cluster with the nearest mean, serving as a prototype of the cluster. It is a widely used technique for exploratory data analysis.  

The challenging issue addressed by the paper is how to support efficient and high-quality clustering while at the same assuring the privacy of individuals to whom the data refer to. The approach described in the paper satisfies differential privacy, a strong privacy notion that aims at approximating the effect of opting-out.  The paper also reports extensive experimental analysis assessing how and why previous algorithms did not perform as well as the proposed technique.  

"The technique developed in the paper is relevant for many different application domains, such as domains requiring data to be clustered around geographical locations. An example of such domains is protection from bio-terrorism in which it is critical to assess unusual high occurrences of diseases symptoms at certain locations,” said Prof. Bertino.”  “Techniques like the one proposed in the paper are increasingly relevant in today big data era as we need to be able to effectively use data while at the same assuring data privacy,” she added.

ACM SIGSAC created the CODASPY conference with the intention of providing a venue for high-quality research in the area of security and privacy, fostering a community of those focused on cyber security. Researchers who participate are seeking new and innovative ways to protect cyber infrastructure, which has become increasingly vulnerable to criminals, spies and predators.

Prof. Bertino also is the Director of the Cyber Center in Purdue's Discovery Park, and Research Director of the Center for Education and Research in Information Assurance and Security (CERIAS).

The winning paper can be viewed at:  

http://dl.acm.org/citation.cfm?id=2857708&CFID=591601884&CFTOKEN=620928