A MATHEMATICAL PROGRAMMING MODEL FOR AUTHENTICATION OF EXAMINEE DURING E-TESTING

In this paper, we study an authentication method in e-Test systems. E-Test is useful for examinees at distant location. However, some drawbacks of e-Test systems are pointed out for practical usage. One of the most critical drawbacks is a difficulty of authenticating the examinees. Many e-Test systems only ask for the password to confirm the authentication at the start of the test. It is difficult to find switching one of examinees to other person during the test at remote location. To overcome this drawback, we consider an e-Test system which authenticates the examinee when taking the test.

Considering the situation that each examinee writes the answers on a pen tablet in e-Test systems, we can use dynamic information of writing of the answers for the authentication without disturbing examinees during the test. The dynamic information is information on the writing process, such as x- and y-coordinates, pen pressure, and the angle and position of the pen, which can be acquired from a pen tablet. Many researchers have studied online signature verification methods using dynamic information. Most of their methods assume that the correct signatures are registered by writers beforehand. They compare the differences between the extracted characteristics from the registered signatures and characteristics of signatures written for authentication. So, in the comparison, the same characters are used. However when we conduct e-tests, it is difficult to collect dynamic information of all possible characters written by the examinees in advance because Japanese language has more than thousand characters.

To reduce the number of characters to be registered, we provide an authentication method using the comparison between characteristics of the different characters. Our proposed method decomposes each character to stroke segments and makes groups of the segments by the same direction of stroke. We calculate the similarity between the dynamic information of the strokes of the answers in e-tests and the same direction group in the registered data. If the similarity is above the predetermined threshold, the examinee is judged to be the correct person. We develop a mathematical programming model to achieve the effective comparison. The proposed method are evaluated using answers given when examinees take an examination by e-testing. The results indicate that our method is applicable for the examinee authentication.