Many important high-stakes decisions—college admission, academic performance evaluation, and even job promotion—depend on accurate and reliable scores from valid large-scale assessments. However, test takers sometimes behave aberrantly (e.g., cheating and guessing), which can undermine the effectiveness of such assessments in yielding accurate, precise information of examinees’ performances. Given this context, the purpose of this research is to create, develop, investigate, and test new statistical approaches that would jointly integrate bio-information (i.e., data) collected from eye-tracking machines with item responses and response time data, which later could be used to detect fraudulent testing behaviors in large-scale testing scenarios. In the current study, three negative binomial models will be introduced for modeling visual fixation counts of test takers solving a set of items. Also, a trivariate joint modeling approach will be introduced that integrates item responses, reaction time and visual fixation counts. This joint model is a cast as a multilevel model in which the structural relationship among the responding accuracy, the working speed and the attention would be manifested. Finally, extensions of using these models for aberrant behavior detections would be discussed. A Markov chain Monte Carlo estimation method is implemented for parameter estimation.