Recent developments have made two-party secure function evaluation (2P-SFE) vastly more efficient. However, because they make extensive use of cryptographic operations, these protocols remain too slow for practical use by most applications. The introduction of Intel’s Software Guard Extensions (SGX), which provide an environment for the isolated execution of code and handling of data, offers an opportunity to overcome such performance concerns. In this paper, we explore the challenges of using SGX to achieve security guarantees similar to those found in traditional 2P-SFE systems. After demonstrating a number of critical concerns, we develop two protocols for secure computation in the semi-honest model on this platform: one in which both parties are SGX-enabled and a second in which only one party has direct access to this hardware. We then show how these protocols can be made secure in the malicious model. We conclude that implementing 2P-SFE on SGX-enabled devices can render it practical for a wide range of applications.