LogAnomaly: Unsupervised Detection of Sequential and Quantitative Anomalies in Unstructured Logs

Authors: Weibin Meng, Ying Liu, Yichen Zhu, Shenglin Zhang, Dan Pei, Yuqing Liu, Yihao Chen, Ruizhi Zhang, Shimin Tao, Pei Sun, Rong Zhou

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our evaluation on two public production log datasets show that Log Anomaly outperforms existing log-based anomaly detection methods.
Researcher Affiliation Collaboration 1Tsinghua University 2University of Toronto 3Nankai University 4Huawei 5Beijing National Research Center for Information Science and Technology (BNRist)
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks (no clearly labeled algorithm sections or code-like formatted procedures).
Open Source Code No The paper does not provide concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described.
Open Datasets Yes We conduct experiments over the BGL dataset [Oliner and Stearley, 2007] and the HDFS dataset [Xu et al., 2009]
Dataset Splits No The paper states, 'we leverage the front 80% (according to the timestamps of logs) as the training data, and the rest 20% as the testing data.' This describes a train/test split but does not explicitly mention a validation set or its details.
Hardware Specification Yes We conduct all the experiments on a Linux server with Intel Xeon 2.40 GHz CPU and 64G memory.
Software Dependencies Yes We implement Log Anomaly and Deep Log with Python 3.6 and Keras 2.1.
Experiment Setup Yes The Log Anomaly in our experiments has two LSTM layers with 128 neurons, and the size (step) of window is 20 (1).