Will a fingerprint sensor be Apple’s next hit?
Some months ago, the press reported that AuthenTec was acquired by Apple. In the last few weeks, rumors on a fingerprint sensor embedded in the next iPhone generation have been increasing in the media.
What would the fingerprint sensor be used for? Why do we need one on the iPhone or iPad? To make payments? I don’t think so.
Your finger tips
If you look really closely at your finger tips, you can see very thin lines flowing smoothly all over the tip surface. They are called friction ridges. We have them to increase the friction between our fingers and the object we are holding in our hand. If you look even closer (with a magnifying glass), you will see that these ridges sometimes interrupt abruptly (ridge endings) and sometime bifurcate (ridge bifurcations). The next figure highlights these points:
The distribution of the ridge endings and bifurcations on the finger surface are unique. This uniqueness makes it possible to recognize people through their fingerprints. [Actually, the concept of fingerprint uniqueness is quite controversial in the scientific community, because until now nobody was able to demonstrate it. However, since nobody has ever found 2 identical fingerprints (even 2 identical twins have different fingerprints), the scientific community agrees on their uniqueness.]
Biometrics
Fingerprint recognition is part of a science called biometrics that uses the unique physical and behavioral characteristics to recognize individual people. Face, palm-print, voice, iris, retina, blood vessel shape, ear shape, gait and other physical and behavioral characteristics are examples of biometrics traits that can be used to recognize individuals. Each one of these treats has advantages and disadvantages and some of them are more reliable, secure, easy-to-capture, intrusive than the others. A biometric modality is chosen in place of another depending on the application and the use case. Some applications also combine more biometric modalities (biometrics fusion) to obtain more accurate recognition (for example, voice and face, or fingerprint and iris, and so on).
Identification and verification
Before we go on, we need to make a clear distinction between identification and verification. Sometimes, these two terms are misused and confused. Identification or 1-to-many recognition is used to compare a fingerprint against a (large) set of fingerprints. This is usually used, for example, when you enter the United States: you provide your 10 fingerprints and the system compares them against all the fingerprints contained in a database of criminals or wanted people (the so-called black list). This is done with an Automated Fingerprint Identification System or, AFIS.
Verification or 1-to-1 recognition is used to check if you are really who you claim to be when you try to access a system or a device. Verification will be used in the iPhone. This is important to know, because verification is much easier to perform than identification and allows less constraint during the recognition and the use of lower quality fingerprint sensors.
Verification requires a reference fingerprint (the so called, template) that you have to provide the first time you use the system or device. Then, every time you try to access that same system or device, you will need to provide the same fingerprint to be compared against the template. If the system or the algorithm recognizes that the 2 fingerprints come from the same person, access will be granted, otherwise you will be rejected.
As with any other verification system, the iPhone will have an enrolling phase. As Apple knows how, this phase can be completely seamless for the user. Like Siri, the new fingerprint technology should be able to learn your fingerprints from the first moment you touch the device.
Fingerprint Sensors
Let’s take a look at the most critical component of the entire fingerprint recognition process: the sensor. Recognition is highly dependent on the quality of the captured fingerprint image as in less noise, and better image. The core technology used to manufacture the sensors can introduce noise and errors on the captured fingerprint image, influencing the recognition to such a negative extreme that you could be continuously rejected by the system (false rejection) or somebody else could be granted access to the system instead of you (false acceptance).
There are 2 kinds of fingerprint capture methods: touchless and touch-based fingerprint capture. I pioneered the first category. This capture technique requires a camera with very sophisticated optical lenses and a complex lighting system. The huge advantage of the touchless fingerprint devices compared to the touch-based ones is that, since the finger does not need to touch any rigid surfaces, the skin does not deform and the image captures very rich details that can make the recognition more accurate. Manufacturing costs are the main issue with this kind of devices and they cannot be miniaturized, so touchless fingerprint devices are not suitable for cellphones. If you want to know more about touchless fingerprint capture technology, you can check some of my publications (Touchless Fingerprinting Technology, Advanced Technologies for Touchless Fingerprint Recognition, 3D Touchless Fingerprints: Compatibility with Legacy Rolled Images).
The touch-based fingerprint sensors require the user to touch a rigid surface (the platen). During the touch, the fingerprint can be acquired with different technologies. However, during the touch, the skin deforms making these devices less accurate than the touchless ones.
There are two main types of touch-based technologies: optical sensors and IC’s or CMOS sensors. Optical sensors are more accurate than IC’s, but they have the similar disadvantages of costs and form factor as the touchless devices. If you want to know more about the touch-based optical devices, check here their basic capture technology, known as Total Internal Reflection.
Finally, we get to IC or CMOS (Complementary Metal Oxide Semiconductor) fingerprint devices. They are made in Silicon in a very similar way to a chip or any other integrated electronic circuit. They can be very thin and very small and, their production costs make them very attractive for mobile applications so they seem like the right candidate to be embedded in an iPhone. And indeed, you should expect this kind of sensor in your next iOS device.
The core technologies used to capture the ridge-valley pattern can be different (temperature, capacitive, electrical and so on). The sensing surface is composed of very small square plates (similar to pixels on a computer screen). To be able to capture a decent fingerprint image, these plates should be not larger than 100 µm x 100 µm in size (the pixel size of an iPhone 5 retina display is 78 µm x 78 µm).
You can imagine the sensor platen as a matrix of extremely small electrical capacitors with each element of this matrix measuring the electrical capacity of different points of the skin when touching the sensing surface. A fingerprint image obtained with a fingerprint touch-based CMOS sensor is shown here:
While the advantage of this kind of sensor is their price and form factor (they can also be miniaturized), let’s look at their real drawbacks.
To help you understand this, I will simplify the concepts. As I mentioned, the surface of a CMOS fingerprint sensor is composed of very small electrical capacitors. In reality, the skin of the finger and the metal surface of the sensor behave as a capacitor:
In the picture above, you can see the profile of the skin of the finger composed of ridges and valleys. Each point of the skin and each small plate of the sensor surface (platen) constitutes an electrical capacitor. A capacitor is an electrical component able to accumulate and retain electrical charges. The touch screen of your iPhone or iPad works in a very similar way, but it is less accurate.
When you touch the sensing surface, electrons from your body move through your finger to the sensing surface and accumulate between the finger skin and the sensor surface in correspondence of the finger touch. The counting of electrons in a given point represents the gray level of the final fingerprint image pixel at that point. When you remove your finger from the sensing surface, the capacitors discharge.
Here the first technological issue. An electrical current is basically a flow of electrons moving between 2 points. This flow happens only in electrically conductive materials (mainly in metallic materials). The flow always originates from the point containing a higher number of electrons and terminates at the point containing a lower number of electrons (this is one of the most beautiful principles of modern physics, since every electronic device around you works because of this principle). When you touch the sensing surface of a fingerprint sensor, the electrons move from your finger to the sensor surface. The quantity of electrons arriving to the surface is typically enormous for the small silicon capacitors. Constant usage of the sensor starts to destroy the capacitors and over time, the fingerprint sensor stops to work. To avoid this issue, during the manufacturing process, the sensor surface is covered by insulating material (essentially silicon, processed to become an insulating layer) that protects the metallic surface. The touchscreen of your iPhone is manufactured in the same way. However, the coating layer on the fingerprint sensor surface cannot be too thick otherwise the electrons from your body cannot reach the metallic surface of the sensor to generate a fingerprint image. So, this protecting layer is thin and only used to extend the life of a sensor, but its continuous usage will destroy its surface, making the device useless.
In everyday life, things are even worst. You usually use your hands for different tasks and you usually touch different types of materials. Small portions of the objects you touch accumulate on the skin of your finger. When you touch the fingerprint sensor, you deposit these material on its surface. Additionally, your skin produces sweat (a combination of water and different types of salts) and the sebum (a oily/waxy substance our body produces). When you touch the surface of a fingerprint sensor, the mix of the sweat, sebum and any substance accumulate during your daily activities become a killer combination for the sensor surface that speeds up the destruction of its surface.
Fingerprint sensor manufacturers (including Authentec) never achieved great success in this issue which is why it is not common to see fingerprint CMOS devices on laptops, cars, building front doors or credit cards. Most of the devices you see are optical sensors that require less maintenance and last longer.
If you search for the specifications of a CMOS fingerprint device, you will find a number representing the lifetime of a device. That number is expressed in number of touches (before it completely dies). That number is provided in ideal conditions of usage and in a normal operating environment of temperature and humidity. But remember where you normally use your iPhone. You keep it in your “dirty” pockets, you leave it on different surfaces, and in humid and hot or cold and dry environments. Sometimes water drops on it or you forget it in your car under the sun. All these factors stress the working conditions of the sensor surface and contribute to speeding up its decay process.
Unfortunately there is no existing solution to this. Manufacturers can only try to make the fingerprint sensor last longer, but sooner or later that device will stop working properly. This is also why Apple cannot provide a fingerprint sensor for payments. And if they do, they are making a huge mistake, because the surface destruction process explained above introduces the most dangerous problem in fingerprint recognition: false acceptance, when after a while somebody else can be granted access to your device.
Companies like Motorola, Fujitsu, Siemens, and Samsung have tried to integrate fingerprints in their laptops and handheld devices, but they have all failed because of the poor durability of the sensing surface.
For you, this means that a fingerprint sensor on your phone will break after a while. How long after you buy it? Well, that will depend on where you live, how you use it, where you use it, how careful you are with it, and how clean your hands are.
Geppy
Geppy Parziale has 15+ years of professional experience in pattern recognition and machine vision. He spent most of his researches in fingerprint recognition and biometrics. He pioneered a novel touchless fingerprint technology trying to overcome the major drawbacks of the legacy fingerprint touch-based devices. He also patented and developed a 10-print capture sensor for the National Institute of Justice, Federal Bureau of Investigation, Department of Defense and Department of Homeland Security within the Fast Capture initiative winning a grant of $3.5 millions. He created and sold two sophisticated and highly-accurate fingerprint recognition algorithms based on a novel pattern matching technique that do not use minutiae points. He is currently working on a novel fingerprint recognition algorithm optimized for mobile devices. Geppy also shares his extensive expertise in iOS training classes and specialized consulting also for biometrics.
