Privacy-Preserving Implementation of Local Search Algorithms for Collaboratively Solving Assignment Problems in Time-Critical Contexts

K. Schütz, C. Schütz, S. Jaburek
Schu23d (2023)
Proceedings of the IEEE 2023 Congress on Evolutionary Computation (CEC 2023), Chicago, IL, U.S.A., July 1-5, 2023, IEEE Press, 10 pages, ISBN 979-8-3503-1458-8, DOI: 10.1109/CEC53210.2023.10253978, 2023.

Abstract (English)

Solving real-world optimization problems often requires collaboration among multiple stakeholders. In air traffic flow management, for example, airlines must work together to prioritize individual flights in cases of reduced capacity in the air traffic network. However, when diverse parties are required to share sensitive information to collaboratively conduct optimization, trust becomes an issue. To alleviate those issues, privacy-preserving computation can be utilized to protect the confidential information of participants, which comes with a trade-off in terms of runtime performance. In time-critical contexts, privacy-preserving implementations of deterministic optimization algorithms may not be able to produce a result before the deadline. In this paper, we investigate the effectiveness of using variants of local search algorithms for the search of solutions to an optimization problem in conjunction with multi-party computation for the evaluation of those solutions. We argue that the proposed method using local search algorithms achieves good results in terms of the quality of the found solution while considerably reducing the run time with respect to a privacy-preserving deterministic solution.