A Hybrid Approach to Authentication of Signature Using DTSVM

Abstract

Signature verification is a widely developed area of research for authentication. A biometric method is used for identification and verification. A unique characteristic of a human like palm, iris, voice, fingerprints, etc. are being used for authentication. Generally, in examination, banking, and any other transcation, two types of signatures are used: (a) handwritten and (b) using any digital device like stylus. Signature verification is the most accepted technique to overcome the problem of forgery from the signature. The main aim of our paper is to provide the authentication of signature using support vector machine technique. As we all know that SVM has many different kinds of function used as kernel such as linear kernel, radial basis kernel, Gaussian kernel, etc. For all these kernels, various types of parameter selection algorithm are available. In this research, we propose a hybrid algorithm, i.e., decision tree support vector machine (DTSVM) for multiple class classification. Using the decision tree algorithm, our DTSVM effectively overcomes the forgery factory from the signature with respect to other effective techniques. The previous experimental results explain that the proposed algorithm is able to find the significant results for skilled forgery in terms of false acceptance ratio, false rejection ratio, and equal error ratio and has better classification accuracy as compared with other algorithms applied.