Measure cross-domain transfer-learning savings

Measure the reduction in Digital Twin retraining effort achieved by transfer learning when BALLAST is moved from the SWaT potable-water process to a different industrial process such as wastewater treatment or a power plant.

Background

All reported experiments use the SWaT testbed, while the proposed commissioning procedure is intended to support deployment at facilities without an existing historical dataset. The paper suggests that cross-domain transfer learning could reduce the amount of retraining required for a new facility.

The magnitude of that reduction is not established, so the practical data and commissioning costs of transferring BALLAST across industrial domains remain unresolved.

References

All results come from SWaT. Power grids and chemical plants need their own physics models, and although cross-domain transfer learning should cut the retraining effort substantially, we have not measured that reduction and do not claim a figure for it.

— BALLAST: Physics-Aware Moving Target Defense for Industrial Control Systems Using a Digital Twin Safety Oracle  (2610.18311 - Mogollon-Gutierrez et al., 13 Oct 2026) in Section 5.7, subsection “Single-site evaluation”