Between the first anomalous signal and the physical damage there is a window. We want to measure it, score it in real time, and make it a first-class security metric.
The window is the product
For a cyber-physical system, the interval between the earliest reliable signal of an attack and the first physical consequence is the entire design budget for mitigation. Detection at T—when the collision is already happening—is a naming error pretending to be a defense.
The Cyber-Physical Threat Horizon asks: how long is that window, in practice, across the signals a real system has? Not in a paper's idealised threat model, but for vehicles, drones and industrial robots under real workloads.
Which signals stay informative
Early results from staged scenarios suggest the usable signals change character as the attack progresses: network anomalies fade quickly, sensor timing drift persists longer, and unexpected command-sequence structure is the most durable signal of all. A working horizon score will need to be a composition, not a single detector.
A metric we want the field to adopt
Deliverables include a warning-time benchmark across simulated attack scenarios and the formal definition of the horizon score. Our goal is modest and ambitious at once: that 'expected warning time' joins boot time and latency as metric people quote when they buy or build a robot.