In recent years, encryption algorithms have become widely adopted as a trusted means of securing data transmission, effectively safeguarding against unauthorized access. However, providing raw encrypted traffic directly to researchers can still pose privacy risks, making it essential to implement desensitization measures before sharing such data. The main challenge is to desensitize the encrypted traffic in a privacyprotection way while preserving as much of the original traffic characteristics as possible to ensure the downstream tasks run properly. In this paper, we propose MADE, an encrypted traffic desensitization framework based on Masked AutoEncoder (MAE). This framework transforms the traffic desensitization task into a classifier-guided image reconstruction problem, using supervised signals to achieve state-aware traffic reconstruction, which ensures the utility as well as the safety of the desensitized traffic for downstream tasks. Moreover, MADE is highly scalable, permitting users to customize which data attributes to preserve in the traffic being processed, simply by modifying the classifier. Experimental results reveal that our proposed desensitization framework not only outperforms baseline methods but also exhibits high applicability across various scenarios.

Open-set Semi-supervised Learning (OSSL) holds a realistic setting that unlabeled data may come from classes unseen in the labeled set, i.e., out-of-distribution (OOD) data, which could cause performance degradation in conventional SSL models. To handle this issue, except for the traditional in-distribution (ID) classifier, some existing OSSL approaches employ an extra OOD detection module to avoid the potential negative impact of the OOD data. Nevertheless, these approaches typically employ the entire set of open-set data during their training process, which may contain data unfriendly to the OSSL task that can negatively influence the model performance. This inspires us to develop a robust open-set data selection strategy for OSSL. Through a theoretical understanding from the perspective of learning theory, we propose Wise Open-set Semi-supervised Learning (WiseOpen), a generic OSSL framework that selectively leverages the open-set data for training the model. By applying a gradient-variance-based selection mechanism, WiseOpen exploits a friendly subset instead of the whole open-set dataset to enhance the model's capability of ID classification. Moreover, to reduce the computational expense, we also propose two practical variants of WiseOpen by adopting low-frequency update and loss-based selection respectively. Extensive experiments demonstrate the effectiveness of WiseOpen in comparison with the state-of-the-art.