Table 1 Threat model of privacy inference attacks in the training phase, YÂ (Yes), NÂ (No)
From: Threats, attacks and defenses to federated learning: issues, taxonomy and perspectives
| Â | Knowledge | Ability | Auxiliary data | ||||
|---|---|---|---|---|---|---|---|
Model structure | Weights | Gradients | Train model | Design model | Modify update | ||
Server | Y | Y | Y | N | Y | Y | N |
Eavesdropping | N | Y | Y | N | N | N | N |
Workers | Â | Â | Â | Â | Â | Â | Â |
 k = 2 | Y | Y | Y | Y | N | Y | Y |
 k > 2 | Y | Y | N | Y | N | Y | Y |