Abstract:A cyber–physical system is a mechanism controlled or monitored by computer-based algorithms, tightly integrated with the internet and its users. Cyber–physical systems such as smart grid, autonomous automobile systems, medical monitoring, process control systems, robotics systems, and automatic pilot avionics will use physical sensors to produce and collect data. Most of the data contains personal information, which is so called privacy, should be carefully protected. How to protect privacy is now a hot-topic not only in academia but also in industry. Differential privacy has been accepted as the privacy concept due to its concise definition and its simple implementation. However, the interactive model cannot achieve differential privacy without data provider’s timely answers, which means data provider should always be attachable. It is unrealistic to keep data provider online due to the risk of data provider be broken will grow rapidly as time goes by. With today’s differential privacy technology, a non-interactive model remains an open problem. To find an alternative, we consider implant whole dataset into a cloud server to provide all the functions instead of data provider. Nonetheless, once the server is compromised, the privacy of the data cannot be guaranteed. It appears that there should be a strong definition, the cloud server is completely trustworthy, before differential privacy can actually be implemented. An intuitive thought to improve this situation is to only upload encrypted datasets. Then, the server could be semi-honest or even fully malicious. Homomorphic encryption can make the encrypted dataset operable, but it requires considerable storage space and bandwidth, which are impractical. We realized that order-preserving encryption is a tradeoff between data utility and practicability. Thus, we propose a novel outsourcing differential privacy data release scheme in cyber–physical system. The proposed scheme allows data providers to outsource their datasets to a cloud service provider with low communication cost. Let the cloud service provider be the host that answers the queries from the data evaluator with noisy results. The data providers can go offline after uploading their encrypted datasets, which is one of the critical requirements for a practical system. In this paper, we present a detailed theoretical analysis, including proofs of differential privacy and security. We also report an experimental evaluation on real datasets.

Multi-party Security Computation with Differential Privacy over Outsourced Data

Privacy-preserving machine learning with multiple data providers

Secure and Efficient Outsourcing Differential Privacy Data Release Scheme in Cyber-physical System

Securely Sampling Discrete Gaussian Noise for Multi-Party Differential Privacy

Outsourced Privacy-Preserving Random Decision Tree Algorithm Under Multiple Parties For Sensor-Cloud Integration

MSDP: multi-scheme privacy-preserving deep learning via differential privacy

Private, Efficient, and Accurate: Protecting Models Trained by Multi-party Learning with Differential Privacy

Differential Privacy Enhanced Dynamic Searchable Symmetric Encryption for Cloud Environments.

Privacy-Preserving k-Nearest Neighbor Classification over Malicious Participants in Outsourced Cloud Environments

EPP-DMM: An Efficient and Privacy-Protected Delegation Scheme for Matrix Multiplication

A secure and efficient outsourced computation on data sharing scheme for privacy computing

Crypt$ε$: Crypto-Assisted Differential Privacy on Untrusted Servers

Differential Privacy-preserving Distributed Machine Learning

Analog Multi-Party Computing: Locally Differential Private Protocols for Collaborative Computations

Mitigating Leakage from Data Dependent Communications in Decentralized Computing using Differential Privacy

Achieving Secure and Differentially Private Computations in Multiparty Settings

Computing encrypted cloud data efficiently under multiple keys

A Comprehensive Comparison of Multiparty Secure Additions with Differential Privacy

Differentially Private Distributed Computation via Public-Private Communication Networks

Privacy-enhanced Multi-Party Deep Learning

Efficient and Privacy-Preserving Outsourced Calculation with Multiple Keys Using Chinese Remainder Theorem