Cryptology ePrint Archive: Report 2010/326

Privacy-Preserving Multi-Objective Evolutionary Algorithms

Daniel Funke and Florian Kerschbaum

Abstract: Existing privacy-preserving evolutionary algorithms are limited to specific problems securing only cost function evaluation. This lack of functionality and security prevents their use for many security sensitive business optimization problems, such as our use case in collaborative supply chain management. We present a technique to construct privacy-preserving algorithms that address multi-objective problems and secure the entire algorithm including survivor selection. We improve performance over Yao's protocol for privacy-preserving algorithms and achieve solution quality only slightly inferior to the multi-objective evolutionary algorithm NSGA-II.

Category / Keywords: applications / secure computation, evolutionary algorithms

Publication Info: A short version of the paper has been accepted for publication at PPSN2010

Date: received 2 Jun 2010, last revised 13 Sep 2010

Contact author: daniel funke at ieee org

Available format(s): PDF | BibTeX Citation

Version: 20100913:211841 (All versions of this report)

Short URL: ia.cr/2010/326