Protecting Security-Critical Real-Time Systems against Fault Attacks in Many-Core Platforms

Single-core platforms have been widely used for Many security-critical real-time systems. However, the ever-increasing high-performance requirements demanded by various industries and the advent of serious bottlenecks again increasing the performance of single-core platforms have necessitated the employment of many-core platforms in the design of such systems. This design shift from single to many-core platforms has been accompanied by security issues and has produced emerging security challenges. Fault injection attacks are one of the primary attacks that are used to infiltrate the tasks to reduce the system performance or cause system failures. In this paper, an online security-aware real-time hardware scheduler is proposed and used to avoid fault attacks using the task replication method. In the proposed real-time system, critical tasks and their replicas are scheduled with Least Slack Time first (LST) algorithm independently in the hardware under real-time constraints. Our synthesis and simulation results using Xilinx Vivado 2018.2 indicates that the proposed scheduler guarantees that all critical tasks and their replicas meet their deadlines. The results also show that our scheduler reduces the chance of a successful Fault attack and loss of the final result in critical tasks.

I. Introduction

The widespread use of real-time systems in security-critical applications such as monitoring and controlling physical systems [1], autonomous driving of intelligent vehicles [2], aerospace [3], transportation [4], and sensitive infrastructures such as power grid and water systems [5] has brought some serious Security challenges endangering human life or environmental damages [6]. Therefore, ensuring the security of many-core systems has become a topic for security-critical real-time system research [7], [8]. Figure 1 shows the structure of a typical security-critical real-time many-core system that uses classical security modules. Although there is no easy way to have direct access to these systems and the adversaries face several difficulties in applying traditional security attack methods, there are new ways to infiltrate these systems.