A New Arbiter PUF for Enhancing Unpredictability on FPGA

Table 4

Results of the overall evaluation.

Metrics	FPGA	Conventional Arbiter PUF	2-XOR Arbiter PUF	2-1 Double Arbiter PUF	3-XOR Arbiter PUF	3-1 Double Arbiter PUF	4-1 Double Arbiter PUF	Ideal
Prediction rate [%] (with 1000 training data)	A	86.32	93.50	80.72	85.82	56.47	56.11	50
	B	86.36	95.28	69.52	85.95	57.45	55.60
	C	86.30	95.41	56.64	85.25	56.75	54.73
Uniqueness [%]	A with B	4.72	4.96	41.36	5.96	50.60	50.46	50
	B with C	4.96	5.62	49.70	6.76	51.34	49.86
	C with A	4.44	5.58	48.06	6.32	48.78	49.76
Randomness [%]	A	53.81	6.32	55.19	54.88	55.68	55.67	50
	B	56.53	4.72	31.40	55.05	52.54	54.76
	C	54.00	4.93	50.63	54.96	53.59	54.59
Steadiness [%]	A	0.76	1.43	7.79	1.43	14.11	34.96	0
	B	0.83	1.36	11.22	1.36	10.93	18.99
	C	0.45	0.52	10.05	0.74	10.35	25.85
Cost (# of SLICEs)	—	177	299	303	426	436	577	—

These values are introduced from [17].
The results of [18] shown in Table 3 are the averages of these values.
These values are introduced from [18].