Validation of dynamic signature for identity verification

Machine based identity validation is extremely important to determine the authenticity of documents, for financial transactions, and for e-communication. Recent explosion of frauds have demonstrated the ineffectiveness of password, personal identification numbers and biometrics.

This thesis presents a signature verification technique which is inexpensive, user friendly, robust against impostors and is reliable, and insensitive to factors such as users’ exposure to emotional stimuli. This work has addressed three important issues which are:
• The selection of appropriate features for dynamic and static signatures.
• The suitable classifier for classification of the features.
• The impact of emotional stimuli on the natural handwriting and signatures of the subjects.

The thesis reports a comparison of the dynamic and static signatures and demonstrates that while the dynamic signature technique has a small increase in the rejection of the authentic user (92% compared with 94%), the system is far more discerning regarding the acceptance of the impostors (1% compared with 21%). The work also demonstrates that the use of ’unknown’ as a class reduces the rejection to zero, by putting those into a class who would be asked to repeat the experiment.

This thesis has also studied the impact of emotional stimuli on peoples’ handwriting and signatures and has determined that while the signatures are insensitive to these stimuli, the handwriting is affected by these stimuli. This outcome may be of importance for people who conduct graphology analysis on people because this suggests that while general handwriting is affected by short term emotional changes of people, signatures are a more robust indicator of the person and hence their personality.