We address the problems of network and reconfiguration attacks on an industrial control system (ICS) by describing a trustworthy autonomic interface guardian architecture (TAIGA) that provides security against attacks originating from both supervisory and plant control nodes. In contrast to the existing security techniques which attempt to bolster perimeter security at supervisory levels, TAIGA physically isolates trusted defense mechanisms from untrusted components and monitors the physical process to detect an attack. Trusted components in TAIGA are implemented in programmable logic (PL). Our implementation of TAIGA integrates a trusted safety-preserving backup controller, and a mechanism for preemptive switching to a backup controller when an attack is detected. A hardware implementation of our approach on an inverted pendulum system illustrates how TAIGA improves resilience against software reconfiguration and network attacks.

No abstract