21.02.2013 Views

Advances in Fingerprint Technology.pdf

Advances in Fingerprint Technology.pdf

Advances in Fingerprint Technology.pdf

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

is 1/2 that the identification is valid; and if three <strong>in</strong>dividuals, the chance is<br />

1/3; etc. K<strong>in</strong>gston’s calculation answers the follow<strong>in</strong>g question:<br />

Given a f<strong>in</strong>gerpr<strong>in</strong>t from an unknown source, and assum<strong>in</strong>g an <strong>in</strong>dividual<br />

is selected randomly from among all persons who actually have the f<strong>in</strong>gerpr<strong>in</strong>t<br />

pattern on their f<strong>in</strong>gers, what is the probability that this one <strong>in</strong>dividual<br />

made the f<strong>in</strong>gerpr<strong>in</strong>t?<br />

The situation is analogous to hav<strong>in</strong>g all N persons who can make the<br />

pr<strong>in</strong>t present <strong>in</strong> a closed room. One of the persons <strong>in</strong>side is randomly<br />

selected, and we ask for the probability that this person is the actual source<br />

of the pr<strong>in</strong>t. We can but note that the person is one of the possible sources<br />

and that the probability is 1/N that we have the correct person. The frequency<br />

of the f<strong>in</strong>gerpr<strong>in</strong>t determ<strong>in</strong>es the magnitude of N, the number of persons <strong>in</strong><br />

the room.<br />

Contrast this situation with one <strong>in</strong> which the <strong>in</strong>dividuals with<strong>in</strong> the room<br />

are selected randomly with respect to f<strong>in</strong>gerpr<strong>in</strong>t type, and where we test the<br />

<strong>in</strong>dividuals to determ<strong>in</strong>e if they could have actually made the evidence pr<strong>in</strong>t.<br />

The number of persons <strong>in</strong> the room (N) now represents a population of<br />

suspects to be tested us<strong>in</strong>g the evidence pr<strong>in</strong>t. If f<strong>in</strong>gerpr<strong>in</strong>ts of an <strong>in</strong>dividual<br />

<strong>in</strong> this suspect group match the evidence pr<strong>in</strong>t, what is the significance of<br />

this f<strong>in</strong>d<strong>in</strong>g? This question parallels the practice of f<strong>in</strong>gerpr<strong>in</strong>t comparison,<br />

whereas K<strong>in</strong>gston’s does not. K<strong>in</strong>gston had assumed that his suspect has been<br />

selected solely on the basis of correspondence with the f<strong>in</strong>gerpr<strong>in</strong>t. Rarely<br />

would this be the case; most often, identification by a partial f<strong>in</strong>gerpr<strong>in</strong>t<br />

would be a nearly <strong>in</strong>dependent event. The comparison would be used to test<br />

a few possible suspects rather than to def<strong>in</strong>e the suspect group. When many<br />

suspects are screened by means of the f<strong>in</strong>gerpr<strong>in</strong>t, the chances of false association<br />

rise, as po<strong>in</strong>ted out by Amy. Only <strong>in</strong> the hypothetically absurd<br />

extreme, <strong>in</strong> which the entire population of the world is screened, is K<strong>in</strong>gston’s<br />

calculation valid.<br />

To answer the appropriate question posed above, one must compare the<br />

chance of the evidence occurr<strong>in</strong>g under two hypotheses:<br />

H 1: that the <strong>in</strong>dividual <strong>in</strong> fact made the pr<strong>in</strong>t<br />

H 2: that another (random) <strong>in</strong>dividual made the pr<strong>in</strong>t<br />

Under H 1, it is certa<strong>in</strong> that the pr<strong>in</strong>t would match the <strong>in</strong>dividual. Under H 2,<br />

the probability is the frequency of <strong>in</strong>cidence multiplied by the number of<br />

attempts made to compare the pr<strong>in</strong>t. A likelihood ratio of these two probabilities<br />

gives the relative support of the evidence for the two compet<strong>in</strong>g<br />

hypotheses. 45

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!