A Qualitative Security Analysis of a New Class of 3-D Integrated Crypto Co-processors

Abstract

3-D integration presents many new opportunities for architects and embedded systems designers. However, 3-D integration has not yet been explored by the cryptographic hardware community. Traditionally, crypto coprocessors have been implemented as a separate die or by utilizing one or more cores in a chip multiprocessor. These methods have their drawbacks and limitations in terms of tamper-resistance, side-channel immunity and performance. In this work we propose a new class of co-processors that are “snapped-on” to the main processor using 3-D integration, and we investigate their security ramifications. These 3-D co-processors hold many advantages over previous implementations. This paper begins with an overview of 3-D integration and its prior applications.We then outline security threat models relevant to crypto co-processors and discuss the advantages and disadvantages of using a dedicated 3-D crypto co-processor compared to traditional, commodity, off-chip crypto co-processors. We also discuss the performance improvements that can be gained from using a 3-D approach.