Table 2 The computational overhead of GVSA for one training round with varying user numbers.

From: Group verifiable secure aggregate federated learning based on secret sharing

 

Key distribution (ms)

Secret sharing (ms)

Model training and masking (ms)

Aggregation (ms)

Verify (ms)

100

40.83

1.54

625.19

2.93

0.19

200

41.40

3.19

627.33

6.51

0.22

300

42.76

5.38

623.13

9.37

0.18

400

43.37

7.27

631.65

11.54

0.19

500

45.01

9.13

625.82

18.91

0.21

Back to article page