Efficient Subspace Updating Using Randomized Tensor Sketch for Online Network Security Monitoring

Jiuzhen Zeng, Laurence T. Yang, Chao Wang, Junjie Su, Yajun Wang

Abstract

In the task of online network security monitoring for the next generation internet, the network anomaly detection is an important and powerful technique. However, current methods face issues in terms of accuracy and efficiency. This paper proposes a novel framework named as ESU-RTS that constructs the randomized tensor sketch, upon which a highly effective and efficient subspace updating strategy is developed for the network anomaly detection. In the framework, a randomized tensor container is generated to store the historical and recent traffic data, thereby approximating the streaming data tensor efficiently. On the basis of this container, a compact randomized tensor sketch is constructed via a randomized compression with power iterations, which allows us to fully capture the essential traffic information in a low-overhead way. The efficient subspace updating is then achieved via the economic QR decomposition on the randomized tensor sketch. With the updated subspace, the real-time and accurate anomaly detection can be realized by comparing each outlier score. To evaluate ESU-RTS, extensive experiments are designed and performed on the real-world traffic trace datasets. The results verify ESU-RTS's accuracy and efficiency over the well-established baseline algorithms.
BibTeX
@article{zeng2026efficient,
  author = {Jiuzhen Zeng and Laurence T. Yang and Chao Wang and Junjie Su and Yajun Wang},
  title = {{Efficient Subspace Updating Using Randomized Tensor Sketch for Online Network Security Monitoring}},
  journal = {IEEE Trans. Network Science \& Engineering},
  year = {2026},
  doi = {10.1109/tnse.2025.3633270},
}

Loading related content...

Figure