Assessment is an important part of a teaching process. It is not only an indicator of student learning outcome but also a main learning activity that provides diagnostic feedbacks, tutorials and an opportunity for students to achieve better outcome. Two key issues relating online assessment are addressed in this article: fairness and quality. By fairness we mean that the score has to be an accurate reflection of student's achievement and by quality we mean that objectives of each study topics have to be achieved through the assessment mechanism. Online assessment tasks are usually carried out in a non-supervised environment to meet the need of flexible delivery. Also, as a practice of student centered learning strategy, students are allowed to carry out the assessment tasks in team. Therefore, to realistically reflect student's individual achievement and to avoid plagiarism, the online assessment tasks must be individualized. Most of the popular online testing systems support questions of type: multiple choice, multiple answer, text match and random number with simple arithmetical evaluation. Obviously the question types listed above are far from satisfactory in testing the understanding of complicated mathematical concepts. This implies that teaching quality will be severely affected by simply using the existing system tools in online testing. In this article we discuss the possibility of using an existing online testing system with functions of generating random numbers and limited scientific evaluation power, such as evaluation of exponential, logarithmic and trigonometric functions, to design online assessment tasks which will ensure equal or better teaching quality than the traditional pen and paper assessment.