SGX-based Users Matching with Privacy Protection

Abstract

For users who rely heavily on social networks for entertaining or making friends, sensitive information such as ages, incomes and addresses will be stored in a database without protection. While many companies try their best to protect user privacy, data breaches still happen, resulting in the loss of millions or billions of dollars and the faith of their customers. Therefore, we propose a solution that guarantees the confidentiality and integrity of information while preserving the ability to perform matching over encrypted values. Our solution is built on homomorphic encryption with secure hardware enclaves such as Intel SGX. Our solution resolves challenges such as performing user profile matching on encrypted values without revealing any information to anyone. With the help of multiple servers, user privacy can be protected as long as at least one server is honest and the guarantee of secure hardware makes the secret unlikely to be revealed. Furthermore, a prototype of our system is implemented to measure its performance. The performance analysis and security analysis show the feasibility of our proposed protocols.