A partial homomorphic encryption scheme for secure design automation on public clouds

Abstract

The massive compute power offered by public clouds such as Amazon Web Services EC2 provides for a new paradigm in cost-effective and highly scalable parallel deployment of Electronic Design Automation (EDA) tools. Even though the advantages are myriad, customers perceive an inherent security risk in exposing their IP to the cloud. In this work, we start by outlining the shortcomings of established encryption techniques for use in public machines. Using the example of electromagnetic simulation, we show how the IP (layout and technology) may be reverse-engineered from the Green's function matrix by utilizing the Multi Dimensional Scaling approach. We then propose encryption schemes to defeat such sophisticated hacking attempts using principles of homomorphic encryption, while enabling scalability and computational benefits of public clouds.