Information about how threshold influence on FAR can be found in IDKit documentation in section "4.8 Far and Threshold".


Since IDKit version 2.70, similarity score normalization and consolidation has changed. Until then the score range was <0,20000> and default score was 12400. The new range is <0, 10000> and default threshold is set to 34 which roughly correspond to FAR = 10^(-5).


Following tables shows the mapping between similarity score and FAR (False Accept Rate), FRR (False Reject Rate) and LLR (Log 10 of Likelihood Ratio).


IDKit v2.60 (score <0,20000>, default similarity threshold 12400)

SCORE
FAR
FRR
LLR
8341
0.1
0.0039
-1.6908
8850
0.01
0.0067
-0.0375
9386
0.001
0.0105
0.6221
9978
0.0001
0.0164
1.2481
10627
0.00001
0.0241
1.9566
11361
0.000001
0.0340
2.8017
11479
0.0000001
0.0361
2.9441


IDKit v2.70 (score <0,10000>, default similarity threshold 34)


SCORE
FAR
FRR
LLR
2
0.063
0.0005
-2.5012
8
0.008
0.0008
-1.7673
19
0.000867
0.0013
-0.4113
34
0.000083
0.0021
0.4584
51
0.00001
0.0030
1.4327
68
0.000001
0.0038
2.0805
85
0.0000001
0.0046
3.4786
102
0.00000001
0.0054
4.3236
119
0.000000001
0.0061
5.1777
136
0.0000000001
0.0067
6.4327


Since version 3.0.7.1 the range has been adjusted to <0, 1000> with default Similarity threshold of 40, which corresponds roughly to FAR = FAR = 10^(-4). The table below shows values of thresholds recommended to use in identification for different FAR for 

IDKit v5.3.2 (score <0, 1000>, default similarity threshold 40)


Probe
Fingers

DB
Size

Enrolled
Fingers

Threshold at
FAR = 1%

Threshold at
FAR = 0.1%

Threshold at
FAR = 0.01%

Threshold at
FAR = 0.001%

Threshold at
FAR = 0.0001%
1
100
1
35
47
59
72
87
1100014457708594
1
10000
1
52
66
80
93
128
1
100
10
42
54
67
80
88
1
1000
10
50
63
76
87
96
1
10000
10
59
73
86
99
128
2
100
2
1
12
24
39
46
2
1000
2
6
21
35
49
56
2
10000
2
14
29
44
61
102
4
100
4
1
10
28
43
60
4
1000
4
1
19
38
53
71
4
10000
4
9
28
47
67
82
10
100
10
1
21
56
87
114
10
1000
10
1
37
78
113
139
10
10000
10
9
54
94
127
142

*FAR of 0.01% can be also expressed as 1:10000 or 0.0001