Automatic Enforcement of Expressive Security Policies using Enclaves
Anitha Gollamudi and Stephen Chong
Proceedings of the 29th Annual ACM SIGPLAN Conference on Object-Oriented Programming Languages, Systems, Languages, and Applications (OOPSLA), October 2016.

Hardware-based enclave protection mechanisms, such as Intel’s SGX, ARM’s TrustZone, and Apple’s Secure Enclave, can protect code and data from powerful low-level attackers. In this work, we use enclaves to enforce strong application-specific information security policies.

We present IMPE, a novel calculus that captures the essence of SGX-like enclave mechanisms, and show that a security-type system for IMPE can enforce expressive confidentiality policies (including erasure policies and delimited release policies) against powerful low-level attackers, including attackers that can arbitrarily corrupt non-enclave code, and, under some circumstances, corrupt enclave code.

We present a translation from an expressive security-typed calculus (that is not aware of enclaves) to IMPE. The translation automatically places code and data into enclaves to enforce the security policies of the source program.