Modern computer systems increasingly integrate NVM (non-volatile memory) as part of main memory, e.g., the Intel Optane Persistent Memory module. While such integration effectively addresses the large capacity demands of modern applications, there is an increasing concern on the security of NVM-based computer servers. While Intel SGX (Software Guard eXtensions) achieves high-level security protection, it faces dramatic performance degradation on NVM-based memory systems. This project aims to address the performance issues in SGX-based secure NVM systems without compromising the security requirements. Given the large global footprint of computer servers that rely on Intel SGX to provide security, this project will have an impact by enabling an efficient and secure operating environment for computer programs. This project also enhances undergraduate and graduate education with outreach to under-represented groups.
By developing hardware-assisted zero-copy page remapping and parallel remapping techniques, it enables secure and low-cost dynamic remapping of insecure pages to SGX memory. By developing compiler-assisted user-controlled paging, it reduces the page reuse distances such that the total number of page remaps between the SGX memory and the untrusted memory can be significantly reduced. By developing cache contents based Merkle tree techniques for secure NVM, it enables the adoption of write-back replacement policy for metadata caches and thus effectively reduces the number of NVM writes. This project enables the delivery of simple yet effective software and hardware components that dramatically enhance performance in secure NVM.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
