An Adaptable System-on-Chip Security Architecture for Internet of Things Applications

Abstract

Modern-day System-on-Chip (SoC) security architectures designed for smart connected devices, such as Internet of Things (IoT) and automotive applications, are often confined by two crucial design aspects: in-field configuration and low overhead. Due to the restrictions posed by these design aspects, it is extremely difficult to develop a robust and adaptable architecture for SoC security policies in IoT and automotive platforms. Security policies, on the other hand, are of critical significance as they implement the confidentiality, integrity, and availability requirements of diverse on-chip security assets. During the complex and often a long life of a system, security requirements evolve, giving rise to the need of adapting security policies. Existing SoC architecture and design flow do not provide the flexibility for easy adaptation of SoC security policies based on emerging threats or security requirements. To address these design constraints and subsequent limitations, a novel security architecture and CAD flow is proposed in this work for efficient implementation of diverse security policies. The adaptable architecture and associated CAD flow enable hardware patching through a reconfigurable security policy engine that can be seamlessly and securely upgraded in-field to address unanticipated attacks and update new security requirements. The infrastructure of the proposed security framework is build with three primary building blocks. First, a centralized Reconfigurable Security Policy Engine (RSPE) is introduced to implement and upgrade policies in-field without comprehensive changes in the architecture. Second, a set of smart security wrappers are developed for efficient extraction of security critical event information and avoidance of communication bottleneck. Third, the on-chip debug instrumentation i.e.  the Design-for-Debug (DfD) infrastructure is employed with minimal modification for extensive access to an arbitrary number of signals of the SoC. A suitable CAD framework is also proposed along with the architecture to systematically implement diverse security policies. The result analysis shows that the architecture provides a high level of adaptability with minimal overhead in terms of power, area, energy, and performance. Hence, the security architecture is highly suited for SoC in IoTs and automotive systems operating in a rigid boundary of performance and energy profiles.