Systemization of Knowledge: Resilience and Fault Tolerance in Cyber-Physical Systems

Abstract

Cyber-Physical Systems (CPS) now support critical infrastructure spanning transportation, energy, manufacturing, medical devices, and autonomous robotics. Their defining characteristic is the tight coupling between digital computation and continuous physical dynamics which enables sophisticated autonomy but also creates highly non-linear failure modes. Small disturbances at sensors, firmware, networks, or physical interfaces can propagate through estimation and control pipelines, producing cascading instabilities that defy traditional single-layer reasoning. This Systematization of Knowledge (SoK) unifies nearly two decades of CPS resilience research into a structured Origin-Layer-Effect (OLE) taxonomy. This taxonomy provides a cross-layer lens for understanding how faults arise, how they propagate, and why unrelated CPS failures often share deep structural similarities. By mapping representative systems including RockDrone, MAYDAY, M2MON, HACMS, Byzantine fault-tolerant control, and learning-based recovery mechanisms onto the taxonomy, we reveal patterns of coverage, persistent blind spots, and recurring pathways of fault amplification. Our analysis identifies four structural gaps that span multiple CPS domains: (1) physical-model manipulation, (2) ML-enabled control without stability guarantees, (3) semantic inconsistencies between formal models and firmware, and (4) inadequate forensic visibility across cyber and physical layers. These insights motivate new directions for resilient CPS design, integrating robust control, runtime monitoring, formal assurance, and system-level visibility.