It's the touch of the fingerprint that makes security more real
by Helén Jansson

In the first of a series of articles on fingerprint and embedded biometric systems technology, Helén Jansson, Marketing Manager of Fingerprint Cards, discusses the emergence of low cost fingerprint biometrics.

For the last few thousand years most types of security have relied on the use of knowledge or tokens to control access or protect against identity fraud. This choice is occasionally reinforced by the use of inspection processes such as passport checking at immigration control points, or by combining the two factors to provide the higher level of identity verification that we have with the PIN code and bankcard. Biometric principles have also been around since the beginning - after all, the Egyptian Pharaohs used fingerprint images as signatures - yet now is the time when technology innovation is beginning to make the difference and the advent of machine-driven or automated biometric recognition offers to make the sometimes illusion of security both more real and affordable.

Secure systems have always experienced a weak spot in the control of individual identity. It might be more accurate to call this a blind spot because secure systems really cannot see or know the individual users of passwords, PIN codes, badges, or cards: they can only assume human identity and trust to human behaviour as being conducive to security. This system blindness is exploited regularly: users of access control systems share keys or passwords; workplace absenteeism goes unrecorded because individuals can clock each other in for work; and the theft of credentials is an even larger problem because whole identities, cards and passwords are stolen or credentials hacked from databases. If the answer is to layer security in order to detect or deter these risks, then one of those layers needs to be biometric, it needs to relate uniquely to an individual.

Biometric products have been available for some time and the number of products becoming enabled with biometrics is growing. The supply and demand cycle here is now sufficiently mature to indicate some general expectations for new security technologies with regard to value for money. To begin with, expectations of performance are rising; biometric systems that work effectively for only ninety percent of any user group are of limited value. Secondly, expectations are for decreasing system costs (either of the components or for the implementation and integration). In Fingerprint Cards' view the only type of biometric capable of riding both these trends at once is fingerprint technology, with a strong preference towards the increased use of systems based on low-cost sensors. Both retina and iris biometrics have been shown to perform very well, but at a much higher hardware cost than fingerprint technology. Without the use of more specialised and costly microphones or cameras then speaker or face recognition biometrics cannot compete on performance: typically their performance is indexed to their cost of deployment.

So, what formula does Fingerprint Cards offer the market? In order to get to grips with the fundamental cost reduction required of biometric systems then there are two key fields of activity. The first is the original design of fingerprint imaging technology components, the sensor hardware and the matching software; the second is the optimised integration of these across different products. The availability of silicon chip sensors, operating to a capacitive principle, is well documented. These sensors offer the low cost route to sensor hardware, and although there are optical competitors these are too large for many product applications such as mobile computing. Silicon sensor technology has become established from a number of vendors, each offering different sensors based on the same fundamental measurement of the difference in capacitance between the peaks and troughs of the fingerprint when the skin is in contact with the sensor surface. All capacitive sensors comprise an array of plates, and each of these creates an electronic signal value when in contact with the fingerprint.

There are technically important differences between all of the competing sensors, but there is a division across two main camps: the direct and the reflective approaches. As the name suggests, direct measurement is focused on the contact area between finger and sensor. Reflective methods (sometimes referred to as active) add to this by pulsing an additional signal through the skin on the fingerprint surface, returning this to the capacitive sensor plates. Fingerprint Cards does this using a drive ring at the edge of the sensor, and the difference made by reflection is felt in the performance. These sensors deliver more distinct fingerprint information across a wider variety of skin types and occasional variations such as wet fingers. The classic problem for silicon sensors is their exposure to Electrostatic Discharge (ESD); something that can quickly disable a sensor, or over time can lead to performance degradation as the available fingerprint data becomes poorer. Fingerprint Cards' reflective method makes it possible to give the sensor surface a protective coating that would act as a barrier to the fingerprint signal in any other capacitive sensor technology, but in Fingerprint Cards' case gives ESD resistance up to 15kV of direct to surface discharge. The same coating is wear resistant up to one million touches.

Within the silicon sensor grouping hardware costs are in close, direct proportion to the size of the sensor circuit. It follows that efforts are being made to reduce this size in the name of competitiveness and this happens in accordance with one of two different principles. The flatbed sensor principle, where the finger is placed and held static on the sensor until the fingerprint pattern is read; or the swipe sensor principle, where the finger is drawn over the sensor to read-off the full finger pattern. Fingerprint Cards has optimised its flatbed sensor (the FPC1010) at 10 x 14mm, being the best size for a variety of fingers and to tolerate positioning and rotational differences, but not so large that unnecessary data is captured for processing. The raw data output from the sensor needs to be processed somewhere, and if this system is battery operated then the amount of raw data will be influential in the implementation of the sensor. The FPC1010 delivers raw data of around 30kB, whereas most competing sensors fall within the range 60-90kB; this efficiency can make a significant different to verification time.

The Fingerprint Cards swipe sensor (FPC1030) measures 2.2 x 10mm, around one fifth the area of the flatbed sensor, and the cost is correspondingly lower. The difficulties with swipe sensors come not from the sensor itself - the FPC1030 uses the same reflective capacitive technology as the flatbed sensor, and it even enjoys a thicker coating giving unrivalled resistance to ESD - instead the challenges are found in what happens to the data from the sensor. Moving the finger saves on a requirement for a larger sensor but it also places a burden on the sensor to read, store and transmit the data captured. Most competitor efforts with swipe sensors are faced with or have encountered bandwidth problems here, usually resulting in artificial restrictions on the speed that the finger can be moved over the sensor. Failure to keep within speed restrictions results in false rejections because data is lost or the memory capacity exceeded. The solution deployed by Fingerprint Cards does not experience this user problem or any reduction in performance. Building on the Company's long expertise in the development of biometric Application Specific Integrated Circuits (ASIC) a co-processor (the FPC2010) has been designed and completed late in 2002 to manage the swipe sensor and to handle efficiently the mathematical calculations required to read the fingerprint pattern for biometric matching. This two-chip system, which includes the fingerprint algorithm, offers a second systems alternative to the established flatbed FPC1010 and its partner ASIC processor, the FPC2000. Choice and flexibility are now more important. The choice of two full systems from Fingerprint Cards services this principle but saves the market customer the time consuming and difficult process of meeting the requirements for optimisation

In any biometric system there are three components: the sensor, the algorithm software, and the host processor. Historically the host has been a PC, either on a desktop, or at some networked point within a system. Swipe sensors place a heavy load on this to handle raw fingerprint data, and in many cases the USB interface is fully tasked to transfer this information to the host. Fingerprint Cards takes the view that such specialised operations are more efficient and more secure if they are handled within the local biometric domain. Since the start of its technology development in the early '90s Fingerprint Cards has been committed to the principle that all systems components should be developed in parallel for optimised performance. This is the second part of the Company's value for money formula. System optimisation accounts for a significant part of the technology's performance in the real world. The embedded systems approach is the most portable, not just because of its literally small size, but because the prior integration work offers rapid cost reduction and shorter times to market. By satisfying the toughest market demands then the technology qualifies more quickly for other possible markets.

What does the future hold for fingerprint technology? Undoubtedly there will be more and wider governmental use of the fingerprint, beyond existing practices in the law enforcement community, and much of this will rely on optical technology to meet image capture standards. With cost as the main barrier to the wider adoption of biometric hardware then all eyes should be on the external forces that will drive down the cost of this. Internal developments such as sensor size reduction face in the correct direction, but the wider context of embedded systems development positions companies like Fingerprint Cards within the fastest growing volume market, that of mobile, electronic devices. Growth in this market could do for biometrics what the home video market did for CMOS camera chip development: diversifying product channels, pushing performance up, and pulling cost of biometric ownership down.