To guarantee the security of the wireless network, effort must be paid to software security as software plays a more and more important role in the development of wireless network intelligentization. Identifying security bug reports (SBRs) from bug repository matters much for reducing security risk of wireless network. Cross-project SBR prediction, which uses a prediction model trained with labeled data from one project to predict another project, has been proposed to eliminate SBRs of software products. While reviewing the previous work focused on cross-project SBR prediction, we find the performance (e.g., Recall, F1-score) of cross-project SBR prediction is too low to the production application. This paper proposes a hybrid sampling approach based on text similarity and uncertainty-sampling. We conduct experiments on ten publicly available datasets. The results show our approach could significantly improve the performance of cross-project SBR prediction. On average, the performance of the classification model can be improved by 34%, 64%, 48%, and 11% in terms of Recall, Precision, F1-score, and AUC, respectively.
