It allows people with biometric enabled passports to skip the manual
process of having a border agent check your passport and instead it
uses facial recognition to verify that the person presenting the passport
is in fact the valid holder of that passport. In 2008 the system allowed a
husband and wife to accidentally switch passports and were able to pass
through the automatic system
the biometrics in passports have a 10% failure rate. A GAO report on
the challenges of implementing biometric border security states there is
a 15% error rate for facial recognition systems as the person ages. As the
holders of the biometric passports age then there will likely be an increase
in failures.
integrated fingerprint recognition system. Although it was not the first,
it is, at the time of writing the most recent. Shortly after the introduction
of the iPhone 5S, the complaints concerning the failure of the fingerprint
recognition to recognise the user of the phone, began to appear. At the
time of writing, reports are that around 20% of users are receiving false
negatives when using the fingerprint recognition system of the iPhone 5S.
problems. Around 12% of the population have fingerprints that cannot be
easily read and a NIST report states that 2% of fingerprints are impossible
to read using existing technology. They could be too old or be engaged in
manual labour so the fingerprints have worn off. Women are also known
to have fainter fingerprint ridges than men and the fingerprint ridges in the
Asian population are also faint. So it could be partiularly difficult to read
the fingerprint ridges of an elderly Asian woman. Fingerprints have been
known to change drastically in a short period of time due to wear from
manual labour or damage such as cuts and burns.
multiple factors.
2. DATA SUMMARISATION ALGORITHM
3. DATA COMPARISON ALGORITHM
This means that the same data collection conditions should be maintained.
In the case of the facial recognition system then, the lighting should be
consistent, the subject should be a consistent distance from the camera so
that the face is a consistent size and the subject should look in a consistent
direction - straight into the camera, ideally.
data collection method - there is no point in attempting to use a feature
that you haven't been able to record and the comparison algorithm - there
is no point comparing features that you haven't extracted.
to make up for faults in the data collection and extraction phases. The
robustness that is needed by the comparison algorithm takes two forms.
Firstly for a positive match it should allow a wider variation of features and
secondly, in the case of a true negative, it shouldn't match. These two goals
are in opposition, as we are more permissive with true positives then the
likelihood of a false positive increases.
negative we perform a testing and calibration phase. It is at this stage that
a large number of mistakes can be and are made when calibrating the
comparison algorithm. The calibration process generally consists of testing
the method with a representative sample of the population that will use
the biometric system. The members of that chosen test sample have a
large impact on the final design of the system as they provide the data
that is used to calibrate it. In an ideal world, you would choose a sample
set of people that perfectly represents the entire human population. In
reality, that is incredibly hard to do. Finding representatives for common
groups of people is fairly easy, for example finding males aged 18-40, but
the difficultly comes when looking for representatives for minorities, and as
the minority represents less and less of the general population then finding
representatives to make up part of the sample becomes increasingly more
difficult.
sample then any physical variation will not be part of the training, and is
therefore unlikely to be accounted for in the final system.
of representative samples of people when testing biometric systems will
greatly help in their accuracy and will allow for a larger coverage of the
population when using biometric systems. Combining multiple biometric
systems, or by giving people the choice of which systems to use. Regularly
updating the biometric database will allow for changes in the person as
they age. In the short term we cannot rely on biometric systems to be
100% accurate, there will always be variation within the population and
until that variation is taken into account at the very start of the design of
the biometric system then there will always be failures.