@inproceedings{wang-etal-2025-towards-multi,
    title = "Towards Multi-System Log Anomaly Detection",
    author = "Wang, Boyang  and
      Zang, Runqiang  and
      Guo, Hongcheng  and
      Zhang, Shun  and
      Cao, Shaosheng  and
      Di, Donglin  and
      Li, Zhoujun",
    editor = "Rehm, Georg  and
      Li, Yunyao",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/test-year-match/2025.acl-industry.8/",
    doi = "10.18653/v1/2025.acl-industry.8",
    pages = "83--91",
    ISBN = "979-8-89176-288-6",
    abstract = "Despite advances in unsupervised log anomaly detection, current models require dataset-specific training, causing costly procedures, limited scalability, and performance bottlenecks. Furthermore, numerous models lack cognitive reasoning abilities, limiting their transferability to similar systems. Additionally, these models often encounter the \textbf{{``}identical shortcut''} predicament, erroneously predicting normal classes when confronted with rare anomaly logs due to reconstruction errors. To address these issues, we propose \textbf{MLAD}, a novel \textbf{M}ulti-system \textbf{L}og \textbf{A}nomaly \textbf{D}etection model incorporating semantic relational reasoning. Specifically, we extract cross-system semantic patterns and encode them as high-dimensional learnable vectors. Subsequently, we revamp attention formulas to discern keyword significance and model the overall distribution through vector space diffusion. Lastly, we employ a Gaussian mixture model to highlight rare word uncertainty, optimizing the vector space with maximum expectation. Experiments on real-world datasets demonstrate the superiority of MLAD."
}