Protecting In-memory Data Cache with Secure Enclaves in Untrusted Cloud

Protecting data security and privacy is one of the top concerns in the public cloud. As the cloud infrastructure is complex, and it is difficult for cloud users to gain trust. Particularly, how to guarantee the confidentiality and integrity of in-memory user private data in untrusted cloud faces big challenges. The in-memory data is typically used for online processing that requires high performance and plaintext access in CPU, therefore simple data encryption is infeasible for in-memory data security protection. In this paper, we propose a secure in-memory data cache scheme based on the memcached key-value store system and leverage the new trusted Intel SGX processors to protect sensitive operations. Firstly, we build a secure enclave and design a trusted channel protocol using remote attestation mechanism. Secondly, we propose a cache server partitioning method that decouples the sensitive key-value operations with enclave protection. Thirdly, we implement a secure client library to maintain the original cache semantics for application compatibility. The experimental result showed that the proposed solutions achieves comparable performance with the traditional key-value store systems, while improves the level of data security in untrusted cloud.