One of the typical gaming disorder is cybersickness. Cybersickness is the condition that occurs during or after exposed by the virtual environment. The increasing of cybersickness symptoms in gamers can lead to the poor health condition. Prior studies in investigating cybersickness employ subjective self-reports questionnaire, i.e., simulator sickness questionnaire (SSQ). However, the objective measurement is required to determine the actual condition of subjects due to cybersickness severity level. Therefore, this paper proposed identification of cybersickness severity level using electroencephalograph (EEG) signals. From the EEG, we extract the best feature such as percentage change (PC) of power percentage (PP) in beta and theta frequency band from pre- to post-stimulation. We found a specific pattern of cybersickness that marked by the sudden decreasing of PPβ during the recording between baseline segment (4 minutes) and the last part (4 minutes) of game playing. Unlike previous studies, this paper proposed the rules-based algorithm i.e. CN2 Rules Induction for identifying cybersickness severity level. This giving ease for medical-expert to determine appropriate diagnosis and treatment towards patients. The classification yields the best accuracy of 88.9% using the CN2 rule induction. It is outperforming other classifiers accuracies such as decision tree (72.2%) and SVM (83.3 %). According to the results, incorporating PC of the PPβ feature with the rules-based algorithm is working well for identifying cybersickness severity level from EEG.