Biometric verification is the verification of a person’s identity through one of his/her biometric data (e.g. biometric signature, fingerprint). Signature verification technologies are used in online banking, bank cheque clearing (offline or image signatures), and credit card purchases (online or biometric signatures) worldwide.
For services in application areas where signature verification is extremely important, such as banking transactions, an innovative 'biometric signature verification system' has been developed that decides whether the signature presented belongs to the actual customer.
This technology has received the first place rankings in many international competitions (SVC2004, 4NSIGCOMP2010, SigComp2011, SigWiComp2013, SigWiComp2015).
This technology uses Dynamic Time Warping (DTW) algorithm to optimally align two signatures of varying lengths and measure their dissimilarity. DTW is a dynamic programming algorithm where the best global alignment is found by comparing all possible alignments between two sequences while avoiding unnecessary computation.
Signatures are collected from the Front End User Interface (UI), subsequently encrypted using the API and stored on the SQL Server as a reference. On a verification page, the Front End UI collects a signature and using the API, calls the Verification Engine to verify the collected signature against reference signatures on the SQL Server.
Biometric signature verification is the most natural solution to the problem of authenticating documents digitally. As the personal signature has always been strongly integrated into our social, legal and commercial lives, biometric signature verification is applied as a universally accepted authentication method in the electronic age.
Signature verification technologies are applied in areas where signature verification is extremely important, such as in banking, credit card transactions, online banking, bank cheque clearing (offline or image signatures), credit card purchases (online or biometric signatures), and authentications in laptops/mobile devices.
A standard verification system uses a threshold to accept or reject a signature, but a global threshold does not work well. For instance, longer and more complex signatures result in higher dissimilarity scores (both for genuine as well as forgeries), while simpler/shorter signatures result in lower dissimilarity scores. We normalize dissimilarity scores using statistical reference datasets so that they can be compared to the global threshold. It is unique and more difficult to forge.