Defense

Summary

In this section, we present all currently supported Model Extraction Defense modules in PyGIP.

Submodules

pygip.models.defense.base

pygip.models.defense.RandomWM(dataset[, ...])

A flexible defense implementation using watermarking to protect against model extraction attacks on graph neural networks.

pygip.models.defense.BackdoorWM(dataset, ...)

pygip.models.defense.SurviveWM(dataset, ...)

pygip.models.defense.ImperceptibleWM(dataset)

pygip.models.defense.atom.ATOM

pygip.models.defense.Integrity