The broader impact/commercial potential of this Small Business Innovation Research (SBIR) project is the enablement of verifiable and confidential computing through research and development that improves the performance and cost of the underlying cryptographic computations. Digital services play a foundational role in varied sectors including communication, education, banking, and others. Without cryptography these services would be impossible to conduct online in a secure and private manner. However, many new and more powerful cryptographic techniques remain out of reach due to their computationally expensive nature. Reducing the cost of these operations enables not just secure communication, but also verifiable and confidential computation, of rapidly growing interest to both the enterprise and government segments for use cases such as privacy-preserving machine learning. The market for verifiable and confidential computing solutions is expected to grow to over one billion dollars in the next five years. This SBIR Phase II project proposes to design and develop a high performance cloud-based system for verifiable and confidential computing. Currently many verifiable and confidential computing techniques can not be brought into production due to their computationally expensive nature. This project performs algorithmic research, hardware design, and software development, that will improve the performance and cost of these operations. In particular, this project optimizes cryptographic algorithms to make them more amenable to GPU-based processing, optimize the software implementation of those algorithms to target GPU architectures, devise methods for reducing the amount of data that needs to be transmitted, and develop a scale-out architecture that dynamically scales capacity as the complexity of uses and number of users grow. With the appropriate design, the project expects to develop a system with over an order of magnitude or more performance than existing designs. Furthermore, by creating a cloud-based infrastructure, these resources can be made available to developers around the world, with low latency and low cost. The expected improvements in performance, cost, and access will enable new use cases that have previously been impractical. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
