Passive OS Identification in Imbalanced Dataset

Identification of the Operating System (OS) by network traffic analysis plays an important role in network security management. With the development of encrypted technologies, the use of machine learning algorithms has gained popularity due to their great identification performance. However, these algorithms often require a large amount of training data and struggle to perform well on imbalanced datasets, which is typical in real network scenarios. To address this issue, this study employs passive identification and resampling techniques to process the dataset, significantly improving the accuracy of the model. To deal with the imbalanced dataset, we introduce a combination of undersampling and oversampling techniques. Then, we apply a deep learning model to accomplish OS identification. Furthermore, we conduct experiments to compare the performance of different resampling techniques. The experimental results show that the proposed method has a better performance compared with other methods.