Abstract. Secure two-party computation enables applications in which participants compute the output of a function that depends on their private inputs, without revealing those inputs or relying on any trusted third party. In this paper, we show the potential of building privacy-preserving applications using garbled circuits, a generic technique that until recently was believed to be too inefficient to scale to realistic problems. We present a Java-based framework that uses pipelining and circuit-level optimizations to build efficient and scalable privacypreserving applications. Although the standard garbled circuit protocol assumes a very week, honest-but-curious adversary, techniques are available for converting such protocols to resist stronger adversaries, including fully malicious adversaries. We summarize approaches to producing malicious-resistant secure computations that reduce the costs of transforming a protocol to be secure against stronger adversaries. In addition, we summarize results on ensuring fairness, the property that either both parties receive the result or neither party does. Several open problems remain, but as theory and pragmatism advance, secure computation is approaching the point where it offers practical solutions for a wide variety of important problems.

We design novel, asymptotically more efficient data structures and algorithms for programs whose data access patterns exhibit some degree of predictability. To this end, we propose two novel techniques, a pointer-based technique and a locality-based technique. We show that these two techniques are powerful building blocks in making data structures and algorithms oblivious. Specifically, we apply these techniques to a broad range of commonly used data structures, including maps, sets, priority-queues, stacks, deques; and algorithms, including a memory allocator algorithm, max-flow on graphs with low doubling dimension, and shortest-path distance queries on weighted planar graphs. Our oblivious counterparts of the above outperform the best known ORAM scheme both asymptotically and in practice.

We design and develop ObliVM, a programming framework for secure computation. ObliVM offers a domainspecific language designed for compilation of programs into efficient oblivious representations suitable for secure computation. ObliVM offers a powerful, expressive programming language and user-friendly oblivious programming abstractions. We develop various showcase applications such as data mining, streaming algorithms, graph algorithms, genomic data analysis, and data structures, and demonstrate the scalability of ObliVM to bigger data sizes. We also show how ObliVM significantly reduces development effort while retaining competitive performance for a wide range of applications in comparison with hand-crafted solutions. We are in the process of open-sourcing ObliVM and our rich libraries to the community (www.oblivm.com), offering a reusable framework to implement and distribute new cryptographic algorithms.

Seprase is a cell surface serine protease that is expressed to high levels by invading human breast carcinoma cells. To investigate the role of seprase in breast cancer, MDA MB-231 human mammary adenocarcinoma cells were engineered to express active seprase to high levels. All cells grow rapidly in cell culture. But differences are discovered when the cells are tested for tumorigenicity, growth, and microvessel density by implantation into the mammary fat pads of female severe combined immunodeficient mice. Control transfectants that do not express seprase grow slowly whereas cells that express seprase to high levels form fastgrowing tumors that are highly vascular. Microvessel density is elevated in tumors of two different lines of seprase transfectants to 146 ؎ 67.4 and 144 ؎ 33.42 vessels/mm 2 as compared with 50.5 ؎ 12.9 vessels/mm 2 for tumors of control-transfected cells that do not express seprase. Sepraseexpressing cells are better able to attract blood vessels and exhibit rapid tumor growth.

Although widespread skeletal dissemination is a critical step in the progression of myeloma, little is known regarding mechanisms that control metastasis of this cancer. Heparanase-1 (heparanase), an enzyme that cleaves heparan sulfate chains, is expressed at high levels in some patients with myeloma and promotes metastasis of some tumor types (eg, breast, lymphoma). Using a severe combined immunodeficient (SCID) mouse model, we demonstrate that enhanced expression of heparanase by myeloma cells dramatically up-regulates their spontaneous metastasis to bone. This occurs from primary tumors growing subcutaneously and also from primary tumors established in bone. Interestingly, tumors formed by subcutaneous injection of cells metastasize not only to bone, but also to other sites including spleen, liver, and lung. In contrast, tumors formed by injection of cells directly into bone exhibit a restricted pattern of metastasis that includes dissemination of tumor to other bones but not to extramedullary sites. In addition, expression of heparanase by myeloma cells (1) accelerates the initial growth of the primary tumor, (2) increases whole-body tumor burden as compared with controls, and (3) enhances both the number and size of microvessels within the primary tumor. These studies describe a novel experimental animal model for examining the spontaneous metastasis of bone-homing tumors and indicate that heparanase is a critical determinant of myeloma dissemination and growth in vivo.