Skip to main content
. 2021 Dec 10;7(50):eabe9986. doi: 10.1126/sciadv.abe9986

Table 4. Results for the case study based on Craig Venter’s data and the Ysearch dataset.

Scenarios include (1) no protection, (2) demographics only, (3) random opt-in, (4) random masking, (5) opt-in game, (6) masking game, (7) no-attack masking game, and (8) one-stage masking game. Only genomic attributes could be masked in this case study.

Searched
suboptimal
strategy
Number of
shared
Y-STRs
Inferred
surname
Confidence
score
Utility Privacy Benefit Loss Payoff Scenarios
recommend
0 50 Venter 0.6688 1 0.5 $100 $75 $25 1
1 49 Venter 0.6688 0.9967 0.5 $99.67 $75 $24.67 /
2 48 Venter 0.6688 0.9933 0.5 $99.33 $75 $24.33 /
3 47 Venter 0.6688 0.9897 0.5 $98.97 $75 $23.97 /
4 46 Venter 0.6688 0.9862 0.5 $98.62 $75 $23.62 /
5 45 Venter 0.6688 0.9825 0.5 $98.25 $75 $23.25 /
6 44 Venter 0.6688 0.9786 0.5 $97.86 $75 $22.86 /
7 43 Venter 0.6435 0.9746 0.5 $97.46 $75 $22.46 /
8 42 Venter 0.5484 0.9706 0.5 $97.06 $75 $22.06 /
9 41 Venter 0.5481 0.9661 0.5 $96.61 $75 $21.61 /
10* 40 Karlsson 0.4662 0.9593 1 $95.93 $0 $95.93 6, 7
11 39 Karlsson 0.4661 0.9544 1 $95.44 $0 $95.44 /
/ 0 / / 0.44 0.00 $44 $0.00 $44.00 2
/ 0 / / 0 1 $0 $0 $0 3, 5
/ 7 Hinze 0.4046 0.5043 1 $50.43 $0 $50.43 4
/ 50 / / 1 0.00 $100.00 $0.00 $100.00 8

*The best strategy in the masking game, found using the greedy algorithm with pruning.