Security and Communication Networks

Research Article

[Retracted] Face Security Authentication System Based on Deep Learning and Homomorphic Encryption

Table 1

Face recognition accuracy.


Data set	Method	FaceNet + Mobi 128-D			Origin FaceNet 128-D			PCA FaceNet 64-D
Data set	Method	0.01%	0.1%	1%	0.01%	0.1%	1%	0.01%	0.1%	1%

LFW	No FHE	82.62	93.47	97.25	84.12	94.62	98.67	84.01	94.56	98.75
	FHE (2.5 × 10⁻³)	83.48	93.38	97.25	84.09	94.62	98.67	84.01	94.56	98.72
	FHE (1.0 × 10⁻²)	81.65	93.42	97.16	83.90	94.60	98.65	83.94	94.62	98.72
	FHE (1.0 × 10⁻¹)	77.72	91.37	97.21	79.12	92.39	98.36	78.16	92.86	98.40

IJB-A	No FHE	21.15	43.87	68.32	23.32	45.76	70.72	20.15	45.97	73.69
	FHE (2.5 × 10⁻³)	21.12	43.94	68.27	23.26	45.80	70.68	20.06	46.02	73.60
	FHE (1.0 × 10⁻²)	20.91	42.78	68.15	22.67	45.69	70.23	19.78	45.93	73.70
	FHE (1.0 × 10⁻¹)	18.67	41.53	67.26	19.61	41.98	69.01	17.48	43.80	72.08

IJB-B	No FHE	24.36	47.06	72.61	25.80	48.26	74.51	24.70	47.83	74.61
	FHE (2.5 × 10⁻³)	24.52	46.97	72.59	25.77	48.27	74.46	24.67	47.86	74.59
	FHE (1.0 × 10⁻²)	24.42	46.96	72.61	25.69	48.20	74.39	24.68	47.80	74.59
	FHE (1.0 × 10⁻¹)	22.68	45.18	71.70	23.68	46.01	72.56	22.52	45.67	73.08

CASIA	No FHE	68.12	82.50	91.16	70.87	84.67	93.25	70.85	84.68	93.40
	FHE (2.5 × 10⁻³)	68.12	82.50	91.16	70.87	84.67	93.25	70.83	84.71	93.41
	FHE (1.0 × 10⁻²)	68.09	82.47	91.12	70.85	84.66	93.23	70.89	84.70	93.37
	FHE (1.0 × 10⁻¹)	68.07	82.45	91.09	70.80	84.65	93.21	70.86	84.68	93.36