seminar on Dynamic Signature Verification System

Dynamic Signature Verification System

ABSTRACT

In this paper we propose a client/server based dynamic signature verification systems for the applications requiring user authentication on the WWW. Both client and server side models are proposed. The signature database and the matching process is implemented through the incremental training algorithm and recall process of neural network based Recognition Engine. Client extracts the dynamic features such as changes in speed, pressure and timing that occur during the act of signing locally. The pattern is an encrypted by client software for security and sent to the server for subsequent processing (i.e. tanning/recall).

Contents

Title                                                                                                    Page No.

           Abstract                                                                                    

1         Introduction                                                                                       1

2        Dynamic signature verification System                                         3

3         Signature Technology                                                                     5                     

4        Proposed client-server model                                                          7

5          Feature Extraction                                                                           9         

6         Comparison Process                                                                       10

7         Advantages of DSVS                                                                       11

           

8          Applications of DSVS                                                                     11

9          DSVS:Strength and Weakness                                                    12

10      Conclusion                                                                                         13       

                                                 

            References                                                                                       14       

A

PAPER

ON

“DYNAMIC SIGNATURE VERIFICATION SYSTEM USING  NEURAL NETWORK”

1. Introduction

Signature verification is a process used to recognize an individual's handwritten signature. Dynamic signature verification technology uses the behavioral biometrics of the handwritten signature to confirm the identity of a computer user. This is done by analyzing the shape, speed, stroke, pen pressure and timing information during the act of signing. Natural and intuitive, the technology is easy to explain and trust. As a replacement for a password or a PIN number, dynamic signature verification is a biometric technology that is used to positively identify a person from their handwritten signature. There is an important distinction between simple signature comparisons and dynamic signature verification. Both can be computerized, but a simple comparison only takes into account what the signature looks like. Dynamic

Signature verification takes into account how the signature was made. With dynamic signature verification it if not the shape or the look of the signature that is meaningful; it if the changes in speed, pressure and timing that occurred during the act of signing. Only the original signer can recreate the changes in timing and X, Y and Z (pressure). A pasted bitmap, a copy machine or an expert forger maybe able to duplicate what a signature looks like, but it is virtually impossible to duplicate the timing changes have in X, Y, Z (pressure). The practiced and natural motion of the original signer would require repeating the pattern shown. There will always be slight variation in a person's handwritten signature, but the consistency created by natural motion and practice over time creates a recognizable pattern that makes the handwritten signature a natural for

biometric identification. Signature verification is natural and intuitive. The technology easy to explain and trust. The primary advantage that signature verification system has over other types of biometric technologies is that signatures are already accepted as the common method of identity verification. This history of trust means that people

are very willing to accept a signature based verification system. Dynamic signature verification technology uses the behavioral biometrics of the handwritten signature to confirm the identity of a computer user. Unlike the older technologies of password and key cards - which are often shared or easily forgotten, lost and stolen – Dynamic signature verification provides simple and natural methods for increased

computer security and trusted document authorization.

2. Dynamic Signature Verification System(DSVS)

DSVS brings handwritten signatures to the information age. The technology uses the behavioral biometrics of a handwritten signature to confirm the identity of a computer user. Enterprise-wide and over the internet-unlike a password, PIN number or keycards which are often shared and could be easily lost ,stolen or forgotten, DSVS

provides increased security and trusted document authorization. DSVS may be used to create application for E-commerce, and this application could include non-refutable documents personally signed electronic documents that can be trusted as temper- proof. DSVS offers a secure and natural solution-today-for private communication, data privacy, access control, document authorization, online shopping, electronic payments, member registration, online banking and more.

DSVS may be implemented as technology that requires integration into a client or server application. Software for client or server communication can be provided and at each client side a digitizing tablet is required. The user provides sample signatures at start up enrollment, creating a biometric template that is sorted in a commercially

available database and on a secured DSVS server. During use, to confirm user’s identity, the encrypted data from a subsequently written signature is compared, at the server, with the secured template. The DSVS server users a powerful recognition engine based on neural network.

ÿ An incremental learning algorithm may be used to take of such deviation in signature patterns over time. Some of the features used are shown in figure 2.2.the technology incorporates a learning function, which automatically absorbs and

reflects natural changes in a signature over time.

ÿ Recognition engine uses the timing of changes in pressure, shape, direction, speed and velocity in the Dynamic signature verification Process, while all other verification

algorithms ignored pressures. Biometric identity data is stored in a secured location and is not transported. Less sophisticated architecture may require that the individual's identity data be transported and stored with each copy of every document or transaction, and that the verification is done at the client side. When the recommended

DSVS client/server architecture is utilized the biometric data is secure. With DSVS the identity data is stored, and verification occurs only on a secure server.

ÿ DSVS offers a real-time one-toone security solution without passwords E for government, legal, medical, banking, access control, Data privacy, personalized document approval, online shopping and more. DSVS is the electronic signature for today's networked and online world.

3. Signature Technology

Signature identification systems analyze two different areas of an individual's signature: the specific features of the signature and specific features of the process of signing one's signature. Features that are taken into account and measured include speed, pen pressure, directions, stroke length, and the point in the time when the pen is lifted from the paper. Signature identification devices also can analyze the" static" image of one's signature. In using the “static” image method, the signature identification device captures the image of one's signature and saves it for feature comparisons to the stored template.

To account for the change in one's signature over time, signature identification  systems adopt to any slight variances over time. The way dynamic signature identification system accomplices revealed is by recording the time, history or pressure, velocity, location and acceleration of the pen each time a person uses the system.

4. Proposed Client -- Server Model

The proposed client and server models are shown in figure 4.1 and figure 4.2 respectively. The client extracts the dynamic features such as changes in speed, pressure and timing that occurred during the act of signing locally. The pattern is then encrypted by client software for security and sent to the server for subsequent processing (i.e. Training/Recall).

The design of signature verification systems requires solutions to five types of problems

      

A. Data acquisition

D. Comparison process

B. Preprocessing

E. Performance evaluation

C. Feature extraction

A static signature verification system receives a 2D image as input from the Camera or scanners. Such a system requires a lot of memory and computing power to process the images. The major algorithmic challenge is the required invariance to the current disposition of the writer: no two signatures are fully identical, even after transformation. A dynamic signature verification system gets its input from a digitizer or other, usually pen based, dynamic input device. The signature is then represented as one or several time-varying signals. In other words, the verification system focuses

on how the signatures being written rather than how the signature was written. This provides a better means to grasp the individuality of the writer but fails to recognize the writing itself. Intuitively, this must be correct, being fully in line with science fiction literature: “the irregularity of the hammer blows used by each artisan followed characteristic patterns to an extent that the maker can be identified without question by sampling that pattern. Collectors developed the method to verify authentically. It’s as definite as an eye print, more positive than any skin-print anomaly,”-Herbert The performance of a signature verification system is generally evaluated according to the error representation of a two class pattern recognition problem, that is, with the type 1 (FRR-false rejection rate) and type 2(FAR-false acceptance rate) errors. as the ideal case (i.e., 0 percent on both errors) is questionable to exist, a choice has to be made depending on the application between one of the two error rates equal to zero or the minimization of the total error, FRR+FAR .For entry systems, the false rejection is the most important; for security systems, the false acceptance is most important.

5. Feature Extraction:

· The acquisition stage provides values of pressure exerted against a measure of time using an instrumented digital pen sensitive to pressure. A localized low pass Sum Filter of order 15 is applied to eliminate frequencies greater than 50 Hz (which could be considered as noise).

· Normalization is then conducted to standardize the values of the pressure extracted between 0 and 1. The feature extraction stage consists of two processes. A new technique of segmentation divides the time series data into specifically defined segments. Characteristics like a shape and curves of the graph, high and low points, stationary points and gradient of the graph are similar ones normalization has been done. The improve segmenting method is based on the algorithm to segment the time data graph into major curves. This is done by calculating the difference of pressure exerted between every two points. If the drop of pressure is more equal to 0.035, then a segmenting point is discovered. The first step verifying a signature is done here. If the amount of segments produced by a test signature is very different from the genuine signature being compared to, the test signature is rejected. In the second process, we use time series modeling with Auto Regressive (AR) technique to calculate the AR coefficient from all the segments. Autoregressive (AR) models have proven to be superiors to Fourier methods due to the ability of AR models to handle short segments of data while giving better frequency resolution and smoother power.

In addition AR methods need only one or more cycle of sinusoidal-type activity to be

Present in the segment to produce good spectral peaks and they also provide the ability to observe small shifts in peak frequencies, which are not easily observed with furrier derived spectra. The AR model coefficient can be easily estimated by solving recursively using Levinson- Durbin or Burg method. These coefficients are the used obtain the power spectral density (PSD) values to represent each segment. Combination of all PSD values from all the segments represents the signature.

The landing and verification stage is made up of a neural network topology known as multi-layer perception or M L P implemented as a part of recognition engine. M L P uses the incremental back propagation algorithm to train the network. Training is equivalent to finding proper weights for all the connections such that a desired output is generated for a corresponding input. Using MLP in the context of a classifier requires all output nodes to be set to 0 except for the node that is marked to correspond to the class the input is from (output equal to 1).

7. Advantages

ÿ Natural and intuitive

ÿ Commonly accepted for authentication

ÿ Less intrusive than iris, fingerprint, etc.

8. Applications of DSVS

Applications of signature identification systems have been slow in their adoption by the financial industry due to the low false rejected rates that banks and other financial institutions required. Although it has been reported that Chase Manhattan Bank was on the first bank to taste a signature identification application. Other applications of

signature identification includes:

ÿ Internal revenue service (IRS) has utilized signature identification in electronically file tax returns.

ÿ Employment Services in England to verify an individual that is claiming benefits.

ÿ Pharmaceutical companies are using it to reduce the overall cost and administration of drug regularly submissions to the FDA.

9. DSVS: Strengths and Weakness

ÿ DSVS has several strengths. Because of the large amount of data present in the signature scan template, as well as the difficulty in mimicking the behavior of signing, signature scan technology is highly resistant to impostor attempts .As a result of the low false acceptance rates (FAR), a measure of likelihood that a user claiming a false identity will be accepted, deplorers can have a high confidence level that successfully

matched users who they claim to be. Signature scan also benefit from its ability to leverage existing processes and hardware, such as signature capture tablets and systems based on the public key infrastructure PKI are popular method for data

encryption. Since most people are accustomed to providing their signatures during customer interactions, the technology is considered less invasive that some other biometrics.

ÿ However, signatures scan has several weaknesses. Signature scan is designed to verify subjects based on the threats of their unique signature. As a result, individuals who do not sign their names in a consistent manner may have difficulty enrolling and verifying in signature scan. During enrolment subjects must provide a series of signatures that are similar enough that the system can locate large percentages of the common characteristics between the enrolment signatures. During verification enough

Characteristics must remain constant to determine with confidence that the authorized

person signed. As a result, individuals with muscular illness and people who sometimes sign with only their initials might result in a higher false rejection rate (FRR), which measures the likelihood that a system will incorrectly reject an authorized user. Since many users are UN accustomed to signing on a tablet, some subject’s signatures may differ to their signatures on ink and paper, increasing the

Potential for false rejection.

ÿ The major problem faced with this technology is differentiating between the consistent parts of the signature and the behavioral parts of the signature that vary

with each signing. An individual’s signature is never entirely the same every time it is

signed and can vary substantially over an individual’s lifetime. Allowing for these variations in the system while providing the best protection against possible

forgers is an apparent hurdle faced by this technology.

10. Conclusion

ÿ Signature identification will continue to develop and improve within the biometric industry because of one major advantage:

public acceptance.

ÿ Even if the individual’s biometric measure remains solely on the card carried by the individual, a considerable level of security and privacy concern exists.

ÿ If the measure were to be stored by the third party, even if only for the purposes of backup, then a much higher level of security and privacy concern exists. A central repository for such biometric identifiers would present opportunities for social control that are the staff of antiutopian novels.

Reference

1. Zhao et al, On-line signature verification by adaptively weighted DP matching, IEICE Trans. Informat. Syst.

2. G. Lorette and R. Plamondon, Dynamic approaches to handwritten signature verification, Computer processing of handwriting, World Scientific, 1990, 21-47.

3. http\www.biomatrics.com

4. Nalwa V.S., Automatic On-line Signature Verification, Proceed. Of the IEEE, 85(2), 1997, 213-239.

5. G. K. Gupta and R. C. Joyce, A Simple Approach to Dynamic Hand-Written Signature Verification, 1995.