Abstract:
"Each person has their own unique signature, which is primarily utilized for personal
identification and validation of important documents or legal transactions. Even in modern
times, many commercial situations, such as check payments or registering at an office, still
depend on a manual review of a single known sample to verify the signature's authenticity.
Verification and forgery detection of these signatures is an important task to prevent the
possibility of theft or fraud. Once the verification of the signature is completed storing the
documents that contains the signature is also important to prevent loss of data, manipulation of
data. Also, many entities such as banks require a traceability of the document signed by any
individual.
To avoid the risk of theft or fraudulent activity, it is necessary to implement a system that is
capable of distinguishing between authentic and fake signatures. To determine whether a
signature has been forged or not, various pre-processing stages are required to be performed
on the raw signature images. The proposed work is based on off-line signature verification
using Siamese neural network and to store the documents and maintain traceability blockchain
was used.
To test the model evaluation matrices such as accuracy and F1 Score were used. The accuracy
of the model changes based on the confidence level used and further explained in the testing
chapter. To evaluate the blockchain component basic unit tests were done to verify the
functionality of the smart contract."
