Automatic Inference of Enclave Placement in LLVM Compiler

Learn about an innovative approach to automating enclave placement in the LLVM compiler framework in this 25-minute conference talk from PriSC 2025. Discover how researchers Wesley B Nuzzo, Mohamed Elwakil, and Anitha Gollamudi from the University of Massachusetts Lowell and U.S. Coast Guard Academy have developed an algorithm that takes non-enclave applications with security policies and automatically infers optimal enclave placement, compiling to Intel SGX enclaves using the OpenEnclave framework. Understand how this implementation at the LLVM IR level enables any LLVM-compatible application to leverage Trusted Execution Environments (TEEs) with significantly reduced programmer effort, addressing the tedious and error-prone process of manual application partitioning while maintaining strong confidentiality and integrity guarantees even against privileged but malicious software.