D.C. - I believe
that the effectiveness of "liveness" detection in a given
fingerprint scanning device is mostly a matter of probability. It
involves associating the relationship between the fingerprint captured
and the object (finger), such that the probability of the two being
discrete objects statistically approaches zero. I doubt you would need
to go through all of the mathematical gyrations with Optel's new
technical approach, since it in and of itself eliminates the possibility
for the two objects to be discrete. Optel proposed approach detects and
scans the object when it comes into contact with the scanning surface.
It emits acoustic waves that may be varied within a specific frequency
range, then it receives the return, or waves reflected back to the
receiver within a specific frequency range (based on the emitted
frequency, object characteristics and surface scattering dynamics). The
received waves are then analyzed and processed to reconstruct the
fingerprint image. Any acoustic waves received that are inconsistent
with those of live tissue are discarded.
W.B. - True, but for purposes of clarification,
you should note the following:
Acoustic waves are mechanical waves, and their behavior depends on
mechanical properties of materials. If the velocities of two kinds of
acoustic waves: longitudinal and shear are known, all so-called Lamé
constants (that describes mechanical properties of material) can be
calculated. The situation is much more complicated (but better from the
point of view of finger recognition), if we have to do this with a
finger lying on the sensor plate. A person's finger has a very
complicated structure: the outer, or dead layer of the finger's skin is
made from a very hard material, called keratin. Keratin is elastic due
to its structure, but has mechanical properties very different from the
properties of water. Yet, it is connected to thicker layer of material (live
skin tissue) that has properties similar to water. The finger's inner
structure is very complicated, containing muscles, blood vessels, bone,
etc. It should be noted that the optical and electrical features of
finger are not very special and can be copied with other materials. (Note:
I have discussed this with others, who have conducted studies on this
subject). In contrast, the mechanical parameters are not easy to copy,
if at all, and the associated behaviors of this structure are probably
impossible to accurately reproduce.
Using our existing prototypes, we have determined that material (and
probably structure) differences are causing strong difference (variance)
in the amplitude of the acoustic signal scattered from the skin lying on
the solid state surface, when compared to other artifacts. Furthermore,
it was also noted that there were significant differences in the "character"
of the signal (this can be shown in FFT, among others). It is easy to
assume that these two points are true, because mechanical properties of
skin are very different from mechanical properties of artificial finger.
However, the situation is much more complicated, since we have no
theoretical description for a phenomena that I am calling "contact
scattering". Moreover, we do not know which parameters are causing
what; the object's material or its structure. Indeed, we were very
surprised at the results of the initial testing with shear waves used in
the prototypes of the solid-state device.
From my point of view, we are only scratching the surface of a large
range of possibilities that might become available through phenomena
such as contact scattering. This will become clearer and many questions
answered as our research continues. Even so, the knowledge we have
acquired through our research to date has enabled us to propose a unique
and powerful new technical approach to fingerprint scanning and
recognition that far exceeds the capabilities of existing conventional
technologies. The concepts and technologies for this new device are
proven. Now, it is just a matter of design and implementation to develop
a device in the optimal configuration. I think, that it will certainly
be possible to create such device quicker, than to determine all the
answers and explanations from a theoretical perspective. And, this
explanation was limited to only those issues concerning the "material
and its structure" recognition.
D.C. - Optel's ability to determine the "living"
state of the object being scanned effectively rules out a cost-effective
approach for using a gel appliqué, as it must not only match the
fingerprint, but 'live' tissue acoustic characteristics as well. This is
especially true when you consider that the latter is the result of
technology and developments that are proprietary to Optel's new
technical approach and, therefore, will be virtually impossible to
acquire without costly research or other less ethical means (which
should also be a costly proposition). To augment the inherent 'live'
tissue test in analyzing the returns from acoustic emissions, Optel's
new approach also has the ability to check for pulse, which it can match
with volumetric changes in the blood vessels that it scans within the
subject's finger.
Finally, and this is something that has only briefly been mentioned
within the last week or so, Optel should also have the ability to check
for biometric changes that are typically associated with an individual's
stress level. Therefore, in select mid- to high-security applications,
Optel's design might also provide an indicator of subject duress by
comparing selected biometric data captured during the subject's initial
enrollment and subsequent entries with the current data. This can easily
be used as a flag to notify security of a potential problem.
W.B.: This is also correct. The ensuing paragraphs
provide a more detailed explanation of our position:
Signal scattered from finger contains information coming not only from
the contact area of finger and sensor surface of the solid-state device,
but also from structures lying deeper in the finger, which will be
delayed. Moreover, it also contains data about the changes in time,
which is exclusive to our application of ultrasound as a scanning
medium.
In our experiments we demonstrated that changes in time caused by blood
flow (pulse) could be used for detection of living fingers, even in the
primitive versions of our software. At this point, this functionality is
working well, allowing the detection of artificial fingerprints on a
thin layer of gel applied to a real finger. Yet, even here, I am sure
that we have just scratched the surface of what might prove to be a wide
range of possibilities that can be detected and applied here:
In earlier testing, we detected something that could only be described
as "global" changes in the signal and were not able to detect
fine changes. As our technology, design and testing evolves, it will
allow us to detect and describe how blood is flowing through the vessels
of the finger. We also note that blood must flow in the form of
three-dimensional waves that have a shape, which is surely a
characteristic that has a high probability of being unique to each
person, much like retinal scans or large vein scans of the hand.
Furthermore, it appears to be a very realistic to believe that we will
also be able to be detect and analyze emotional (behavioral) and/or
other physiological changes that impact the "character of blood
flow" of a given subject. Examples of these include, but are not
limited to, changes caused by stress/duress (behavioral) and illness (physiological).
We also know that acoustic signals reflected from the finger also
contain information coming from the inside of the finger, depending on
the time it is received and the nature of the received wave. However,
since we still have a great deal of research to do in this area, we can
only speculate on the possibilities offered by this particular
capability. For example, we believe it is possible to capture and
analyze information received from the finger's internal structure and,
assuming this is true, it should then be possible to reconstruct images
of that structure. However, we have just begun research in this area, so
we still have many more questions than answers regarding its operation
and potential.
Yet, to truly replicate a person's finger in order to create a viable 'fake',
somebody must have definite answers to these questions and more.
Furthermore, the must have the capability to construct the artificial
finger replicating many of the features in the subjects' fingers in a
cost effective manner. So, this process would need to be created on a
production scale, since a single reproduction of a subject's finger
would almost certainly be cost preclusive. Fortunately, even current
state of the art technology will not even come close to considering such
a construction, or re-construction, as the case may be.
Some might propose that cloning technology might allow for the
replication of a real subject's finger, and that this could be done on a
scale that might make it cost-effective. Yet, even a cloned copy of a
real finger would contain internal and external variations in structure
that could be detected to prevent fraud. Furthermore, for such an
approach to be feasible, it would be necessary to cause the blood flow
in the 'replicated' finger to be exactly as in the real one. In truth,
there are probably even more points that must be considered in order to
realize the possibility of finger replicas, as we have described above,
which only serves to further compound the problem. Such a possibility
does not appear to be a realistic goal for the foreseeable future…
especially at the level of exactitude that would be required.
While we will concede that that our approach and resultant capabilities
may seem a bit "esoteric" now, it is only because very few
considered such a possibility as a viable alternative for a biometrics
application. As a result, very few have conducted research in this or a
related area. Yet, biometrics are only one of the technical applications
for the results of our research. We have many possibilities to consider
in terms of technical applications from a medical perspective, just as
we have from a physical perspective in applications such as
non-destructive testing. Given the findings of our research to date,
such applications appear to be within the realm of possibilities, with
great potential for new gains in efficiency and cost-effectiveness.
After all, we have just scratched the surface in terms of the potential
for discoveries in this new area.
Additional comments: It is important to note that even people
with very poor fingerprints that cannot be detected, resolved, and/or
captured with any existing conventional fingerprint scanner, are viable
candidates for using our proposed fingerprint scanning approach. These
users are capable of generating sufficient unique scattering of sound
waves to allow for the capture and reconstruction of a resolvable
fingerprint image. To date, we have concentrated our discussions on
fingerprint recognition using a conventional imaging approach. However,
future discussions will focus on an innovative new imaging approach
based, in part, on the application of holographic technology.
Finally, it is also important to note that all of the capabilities and
features discussed in Optel's approach to "liveness" detection,
above, are inherent to ultrasound. As such, they do not require the
addition of any new elements or components, other than that which is in
the actual device. Modifications, should they be necessary, would be
limited to software changes for additional signal analyses. In certain
cases, it might also be necessary to improve the amplifiers and filters
to enhance its ability to receive weaker signals, such as those from
internal structures of the finger. In the solid-state version of the
device this should be very easy to do. We are also considering the
development of a more suitable device configuration that will allow for
"comfortable" detection of all finger features. Such devices
might no longer be called "fingerprint" readers, but more
aptly called "finger" or "hand" readers.
