A privacy preserving homomorphic computing toolkit for predictive computation. Issue 2 (March 2022)

• Record Type: Journal Article
• Authors: Zhao, Kaiyang; Wang, Xu An; Yang, Bo; Tian, Youliang; Zhang, Jindan
• Is Part Of: Information processing & management. Volume 59:Issue 2(2022)
• Publication Date: 2022-03
• Subjects: Predictive computation -- Homomorphic encryption -- Data outsourcing and processing -- Data management; Information storage and retrieval systems -- Periodicals; Information science -- Periodicals; Systèmes d'information -- Périodiques; Sciences de l'information -- Périodiques; Information science; Information storage and retrieval systems; Periodicals; 658.4038
• DOI: 10.1016/j.ipm.2022.102880
• Languages: English
• ISSNs: 0306-4573

Abstract

Abstract: Predictive computation now is a more and more popular paradigm for artificial intelligence. In this article, we discuss how to design a privacy preserving computing toolkit for secure predictive computation in smart cities. Predictive computation technology is very important in the management of cloud data in smart cities, which can realize intelligent computing and efficient management of cloud data in the city. Concretely, we propose a homomorphic outsourcing computing toolkit to protect the privacy of multiple users for predictive computation. It can meet the needs of large-scale users to securely outsource their data to cloud servers for storage, management and processing of their own data. This toolkit, using the Paillier encryption system and Lagrangian interpolation law, can implement most commonly basic calculations such as addition, subtraction, multiplication and division etc. It can also implement secure comparison of user data in the encrypted domain. In addition, we discuss how to implement the derivative of polynomial functions using our homomorphic computing encryption tool. We also introduce its application in neural networks. Finally, we demonstrate the security and efficiency of all our protocols through rigorous mathematical analysis and performance analysis. The results show that our toolkit is efficient and secure. Highlights: We use the Paillier encryption system and Lagrange's interpolation theorem to construct a new threshold public key cryptosystem with trapdoor, this can lower the risk of private key leakage and the cost of private key management. We build an outsourced computing toolkit to protect data privacy. The toolkit includes commonly used basic operations, such as addition, subtraction, multiplication, division, comparison, sign bit extraction, equivalence testing etc. We extend the toolkit to design a secure encryption scheme for polynomial function and its derivation, we also discuss its application in the secure active function of neural network. We use rigorous mathematical analysis to verify the security and correctness of our schemes, we also evaluate the efficiency of these schemes through theoretical analysis and implementation.