Abstract: To effectively measure the status of personal network security awareness, this paper proposed an adaptive testing of the network security awareness based on the items response theory (IRT) and multi-stage testing theory. First, a knowledge library model based on "industry, job, and personnel" was designed. According to the model, the calculation method of the item parameters was given. Second, the concept of reaction time was imported to improve the logistic-3-parameter model. Finally, the grouping strategy based on the sliding window and the state evaluation model was given. Results show that the improved adaptive testing performance is 34.9% and 8.7% higher than the traditional model testing and the original logistic-3-parameter model testing, respectively. It provides a reference for the field of testing on network security awareness and computerized adaptive testing technology.