Secure channels in an integrated MPSoC architecture
Haris Isakovic, Armin Wasicek · 2013
Providing security in an embedded system often boils down to solving a trade-off problem between security and performance. Simultaneously, Multi-Processor System-on-a-Chip (MPSoC) devices are in the early stages to increase computational performance, energy and die area efficiency, and reduce the number of physical units in the embedded system design arena. Moreover, MPSoCs enable composing heterogeneous subsystems on a single silicon die which is particularly desirable for large volume embedded devices. However, these benefits come at a price: an increase in the system's complexity. Complexity does not only make the system design process more difficult, but also it renders certain vulnerabilities possible. A solution is to follow well-established architectural principles to reduce complexity and to provide the required level of security. In this paper we demonstrate how the basic architectural principles of the ACROSS MPSoC architecture can be combined with the requirements of standard security techniques (i.e., encryption, authentication) to produce an efficient security solution for MPSoC systems. We propose a security architecture which uses the principles of temporal and spatial partitioning, temporal determinism, and mixed-criticality integration to migrate resource expensive security functions form the application components to a dedicated security component within the MPSoC. This leaves application components with a thin security provider, without any loss of functionality and more local resources at their disposal. Thereby, we deliver a flexible, resource efficient security solution, which highlights the benefits of partitioning MPSoC architectures for security.