The previous article in this series addressed voice recognition as a form of biometric technology. Here, expert Ravi Das moves on to the science of signature recognition as a behavioral biometric technology—how it works and the various factors that can affect its accuracy.
Signature Recognition: How It Works
It should be noted that signature recognition is primarily used for verification purposes only. While in the most theoretical sense it could potentially be used for identification applications, in the real world, it would prove to be very impractical.
This is so for a number of reasons. First, the hand can be greatly affected by genetic factors, various types and kinds of physical features (such as ailments, and the aging process of the hand we all experience at some point in time or another). Second, the signature is very dynamic, and can change very quickly over time, whether the individual has intentions to or not in altering their own signature. Third, unlike the other biometric technologies, there is virtually no permanence or long term stability (unlike the iris or the retina) associated with signature recognition, because of its ever changing nature.
The very first signature recognition devices utilized static variables such as the height, spacing, slope, as well as the various characteristics in terms of the shaping of the letters which are found in the signature. By the mid 1970’s, signature recognition became much more dynamic in the sense that various spatial, pressure, and temporal variables were taken into consideration. These variables included such factors as the downward pressure applied to the pen, the level of pressure at which the pen itself is gripped at, and the angle at which the pen is held while the individual signs his or her name, and even the time it takes for the signature to be completed.
Today’s signature recognition devices can now collect and analyze such variables as speed, acceleration, pauses, and the changes in pressure in which the individual signs their name on the special writing tablet. Neural network technology can also be incorporated with signature recognition which can literally learn the ever so slightest changes and variations in the way an individual signs their name over a pre-established period of time and make the necessary changes to the database.
Signature Recognition technology involves the use of a pen and a special writing tablet, which are connected to a local or central computer for processing and verification. To acquire the signature data during the enrolment process, an individual is required to sign his or her name several times on the writing tablet. It should be noted that the robustness of the Signature Recognition enrolment template is a direct function of the quality of the writing tablet.
A high-quality tablet will capture all of the behavioral variables (timing, pressure, and speed). In contrast, a low-spec tablet may not be able to capture all these variables. There are several constraints to the data acquisition phase. First, a signature cannot be too long or too short. If a signature is too long, too much behavioral data will be presented. As a result, it will be difficult for the Signature Recognition system to identify consistent and unique data points.
If a signature is too short, insufficient data will be captured, giving rise to a higher False Accept Rate. Second, the individual must complete the overall enrolment and verification process in the same type of environment and under the same conditions.
For example, if the individual stands during enrolment, but sits down during verification, the enrolment and verification templates may vary substantially (this is attributable to the amount of support given to the arm). Once the data acquisition phase has been completed, the signature recognition system extracts unique features from the behavioral characteristics, which includes the time needed to place a signature, the pressure applied by the pen to the writing tablet, the speed with which the signature is placed, the overall size of the signature, and the quantity as well as direction of the signature strokes.
With signature recognition templates, different values or ‘weights’ are assigned to each unique feature. These templates are therefore as small as 3kB. One of the biggest challenges in signature recognition is the constant variability in the signatures themselves. This is primarily due to the fact that an individual never signs their signature in the same fashion any two successive times.
For example, the writing slope can switch tangentially left to right (and vice versa), and up and down (and also vice versa); the exact pressure put on the pen can change greatly each and every time the individual has to submit a verification template; and even a reflective light on the surface of the signature recognition capture device can indirectly cause variances in the speed as well as the timing of the signature.
Up Next: The next article will review the strengths and weaknesses of signature recognition, as well as its market applications.
Ravi Das is a Cybersecurity Consultant and Business Development Specialist. He also does Cybersecurity Consulting through his private practice, RaviDas Tech, Inc. He also possesses the Certified in Cybersecurity (CC) cert from the ISC2.
Visit his website at mltechnologies.io