Joint Paper between Aalborg University and BIFOLD at The Web Conference 2026
The ACM Web Conference 2026 took place between June 29th and July 3rd, in Dubai, United Arab Emirates. Since 1994, the conference has been the leading forum for research shaping the future of the web and starting in 2022 transitioned to an ACM event. This year, Volker Markl, BIFOLD Co-director and researcher, co-authored a paper on streaming time series anomaly detection.
Most existing anomaly detection methods are static and target predefined anomaly types, which limit their performance on streaming time series, where data distributions and anomaly formats change over time. To address this challenge, the authors propose DESS, a data-efficient framework for streaming anomaly detection. DESS combines an evolving proxy-generation module that summarizes historical data, a heterogeneous temporal feature-extraction module that captures correlations across multiple semantic levels, and a parameter-efficient training scheme that activates only a subset of parameters. Experiments on real-world data show that DESS outperforms the best baselines by up to 17.53% while reducing training time by up to 64.88%.
Reference: “Evolving Proxy Kills Drift: Data Efficient Streaming Time Series Anomaly Detection,” authored by Qing Wei, Hao Miao, Yan Zhao, Kai Zheng, Bin Yang, Volker Markl, and Christian S. Jensen, https://doi.org/10.1145/3774904.3792440.