Protocols for secure computation allow mutually distrusting parties to compute an arbitrary function of their joint data while preserving privacy of their inputs (to the extent possible), ensuring correctness, providing fault tolerance, and more. Secure computation has been suggested for such diverse applications as privacy-preserving data mining, secure information sharing, privacy in the cloud, and secure database search.

Although feasibility of generic secure computation (for arbitrary functions) was established over 25 years ago, for a long time the perception was that it was hopelessly impractical. Recent results by myself and colleagues indicate that this is no longer true. I will survey several such results, including:
- A recent framework for secure two-party computation that leads to dramatic improvements in      efficiency and scalability compared to what was previously available.
- Application of this framework to the specific problem of private set intersection, which shows that generic protocols can be competitive with "specially designed" protocols for this problem.
- A relaxed (yet meaningful) security definition for security against malicious adversaries that can be achieved 100x more efficiently than in prior work.
- Work showing feasibility and practical efficacy of sublinear-time secure computation.