ePMLF: Efficient and Privacy-Preserving Machine Learning Framework Based on Fog Computing

Table 7

Comparison of computation overhead: model owner and data owners.

Building blocks	Model owner (ms)		Data owners (ms)
	Ours	[5]	Ours	[5]
Secure addition	0.019	0.08	0	0
Secure subtraction	42.6	80.5	0	0
Secure inner product	10.1	79.6	0	0
Secure summation	11.93	1410.23	133.64	306.15
Secure sigmoid function	139.45	231.65	92.96	190.67