CoChat

Impact of log parsing on deep learning-based anomaly detection

Open access
Khan, Zanis Ali, Shin, Donghwan, Bianculli, Domenico, Briand, Lionel C. · 2024 · Empirical Software Engineering

The DOI matches its Crossref record; 46 of the 46 references it cites were found in a registry or an index.

Abstract

Software systems log massive amounts of data, recording important runtime information. Such logs are used, for example, for log-based anomaly detection, which aims to automatically detect abnormal behaviors of the system under analysis by processing the information recorded in its logs. Many log-based anomaly detection techniques based on deep learning models include a pre-processing step called log parsing. However, understanding the impact of log parsing on the accuracy of anomaly detection techniques has received surprisingly little attention so far. Investigating what are the key properties log parsing techniques should ideally have to help anomaly detection is therefore warranted. In this paper, we report on a comprehensive empirical study on the impact of log parsing on anomaly detection accuracy, using 13 log parsing techniques, seven anomly detection techniques (five based on deep learning and two based on traditional machine learning) on three publicly available log datasets. Our empirical results show that, despite what is widely assumed, there is no strong correlation between log parsing accuracy and anomaly detection accuracy, regardless of the metric used for measuring log parsing accuracy. Moreover, we experimentally confirm existing theoretical results showing that it is a property that we refer to as distinguishability in log parsing results—as opposed to their accuracy—that plays an essential role in achieving accurate anomaly detection.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1007/s10664-024-10533-w

Cites

46 of 46 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 An example of making a distinguishable log parsing result indistinguishable by merging templates
Fig. 1 An example of making a distinguishable log parsing result indistinguishable by merging templatesCC BY · PubMed Central
Algorithm 1 Generating an indistinguishable log parsing result from a distinguishable one
Algorithm 1 Generating an indistinguishable log parsing result from a distinguishable oneCC BY · PubMed Central
Algorithm 2 Generating a distinguishable log parsing result from an indistinguishable one
Algorithm 2 Generating a distinguishable log parsing result from an indistinguishable oneCC BY · PubMed Central
Fig. 2 Relationship between TI accuracy and AD accuracy (HDFS)
Fig. 2 Relationship between TI accuracy and AD accuracy (HDFS)CC BY · PubMed Central
Fig. 3 Relationship between TI accuracy and AD accuracy (Hadoop)
Fig. 3 Relationship between TI accuracy and AD accuracy (Hadoop)CC BY · PubMed Central
Fig. 4 Relationship between distScore and AD Accuracy (HDFS)
Fig. 4 Relationship between distScore and AD Accuracy (HDFS)CC BY · PubMed Central
Fig. 5 Relationship between distScore and AD Accuracy (BGL)
Fig. 5 Relationship between distScore and AD Accuracy (BGL)CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

  • Random Forests

    Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538 · cited 2×

    “…learning models, we included two representative traditional machine learning models, namely Support Vector Machine (SVM) (Hearst et al. 1998) and Random Forest(RF)Breiman(2001) 2 sincetheyareknownfortheireffectivenessinanomalydetection tasks on the HDFS dataset (Wu et al. 2023; Jia et al. 2023).”
    Method
  • BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173 · cited 3×

    “…log representation techniques, such as FastText (Joulin et al. 2016), Word2Vec (Mikolov et al. 2013), TF-IDF (Salton and Buckley 1988) and BERT (Devlin et al. 2018), used to convert textual log data into numerical feature vectors for machine learning algorithms, such as Support Vector Machine,…”
    BackgroundMethod
  • Efficient Estimation of Word Representations in Vector Space

    Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013) · arXiv · doi:10.48550/arxiv.1301.3781 · Cited by 26,568 · cited 3×

    “They considered different log representation techniques, such as FastText (Joulin et al. 2016), Word2Vec (Mikolov et al. 2013), TF-IDF (Salton and Buckley 1988) and BERT (Devlin et al. 2018), used to convert textual log data into numerical feature vectors for machine learning algorithms, such as Support Vector Machine…”
    Method
  • Term-weighting approaches in automatic text retrieval

    Salton, G., & Buckley, C. (1988) · Information Processing & Management · doi:10.1016/0306-4573(88)90021-0 · Cited by 9,740 · cited 2×

    “They considered different log representation techniques, such as FastText (Joulin et al. 2016), Word2Vec (Mikolov et al. 2013), TF-IDF (Salton and Buckley 1988) and BERT (Devlin et al. 2018), used to convert textual log data into numerical feature vectors for machine learning algorithms, such as Support Vector Machine…”
    Method
  • Support vector machines

    Hearst, M. A., Dumais, S. T., Osuna, E., Platt, J., & Scholkopf, B. (1998) · IEEE Intelligent Systems and their Applications · doi:10.1109/5254.708428 · Cited by 7,243 · cited 8×

    “In Table 6, except for PLELog, SVM, and RF, the values for all anomaly detection techniques are identical due to the reasons explained in the last paragraph of Section 5.1.”
    MethodInfluential
  • On the Properties of Neural Machine Translation: Encoder–Decoder Approaches

    Cho, K., van Merrienboer, B., Bahdanau, D., & Bengio, Y. (2014) · Proceedings of SSST-8, Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation · doi:10.3115/v1/w14-4012 · Cited by 7,099

  • Individual Comparisons by Ranking Methods

    Wilcoxon, F. (1992) · Springer Series in Statistics · doi:10.1007/978-1-4612-4380-9_16 · Cited by 2,915 · cited 1×

    “To see if the differences between R dst and R ind are significant, we applied the non-parametric L og R obu s t ( F 1 ) Wilcoxon signed rank test (Wilcoxon 1992) for paired samples to the F1 scores of R dst and R ind , for each of the seven anomaly detection techniques and the two datasets.”
    Method
  • DeepLog: Anomaly Detection and Diagnosis from System Logs through Deep Learning

    Du, M., Li, F., Zheng, G., & Srikumar, V. (2017) · Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3133956.3134015 · Cited by 1,835 · cited 12×

    “…would be enough in some cases (e.g., detecting anomalies based on the sequence of template IDs without considering the content of the templates (Du et al. 2017)), correctly identified templates (i.e., templates identical to the corresponding oracle ones) matter when the fixed parts of templates…”
    BackgroundMethodInfluential
  • Drain: an online log parsing approach with fixed depth tree

    He, P., Zhu, J., Zheng, Z., & Lyu, M. R. (2017) · 2017 IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws.2017.13 · Cited by 943 · cited 7×

    “…various elements of each variable part with a special character “ <*> ”; this representation is widely used in log-based analyses (e.g., log parsing (He et al. 2017; Jiang et al. 2008), anomaly detection (Zhang et al. 2019; Du et al. 2017), and log-based testing (Elyasov 2012; Jeong et al.…”
    BackgroundMethodInfluential
  • FastText.zip: Compressing text classification models

    Joulin, A., Grave, E., Bojanowski, P., Douze, M., Jégou, H., & Mikolov, T. (2016) · arXiv · doi:10.48550/arxiv.1612.03651 · Cited by 887 · cited 4×

    “Furthermore, this is the first work to empirically demonstrate the importance of distinguishability after the theoretical framework proposed by Shin et al. (2021).”
    BackgroundMethodInfluential

Show all 46 references

Research works printed but not linked: 7

  1. Ali Abd Al-Hameed K (2022) Spearman’s correlation coefficient in statistical analysis. Int J Nonlinear Anal Appl 13(1):3249–3255
  2. He S, Zhu J, He P, Lyu MR (2020) Loghub: A large collection of system log datasets towards automated log analytics. arXiv:2008.06448https://arxiv.org/pdf/2008.06448.pdf
  3. Huang C, Guan H, Jiang A, Zhang Y, Spratling M, Wang YF (2022) Registration based few-shot anomaly detection. European conference on computer vision. Springer, Springer, New York, NY, USA, pp 303–319
  4. Meng W, Liu Y, Zhu Y, Zhang S, Pei D, Liu Y, Chen Y, Zhang R, Tao S, Sun P et al (2019) Loganomaly: Unsupervised detection of sequential and quantitative anomalies in unstructured logs. IJCAI, ACM, New York, NY, USA 19:4739–4745
  5. Oliner A, Stearley J (2007) What supercomputers say: A study of five system logs. 37th annual IEEE/IFIP international conference on dependable systems and networks (DSN’07), IEEE, IEEE/IFIP International Conference on Dependable Systems and Networks (DSN’07). Edinburgh, UK, pp 575–584
  6. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V et al (2011) Scikit-learn: Machine learning in python. J Mach Learn Res 12:2825–2830
  7. Vaarandi R, Pihelgas M (2015) Logcluster - a data clustering and pattern mining algorithm for event logs. In: 2015 11th International conference on network and service management (CNSM), IEEE, Los Alamitos, CA, USA, pp 1–7. https://doi.org/10.1109/CNSM.2015.7367331 · DOI · Paper page

Cited by

See all 84

Metadata

Authors
Khan, Z. A., Shin, D., Bianculli, D., & Briand, L. C.
Affiliations
  • University of Luxembourg (LU)
  • University of Sheffield (GB)
  • University of Ottawa (CA)
Corresponding author
Khan, Zanis Ali
Venue
Empirical Software Engineering
Published
2024-08-17
Volume
29
Issue
6
Pages
139-139
ISSN
  • 1382-3256 (ISSN-L)
  • 1573-7616
OpenAlex
W4401667316
Cited by
49
References
42
Citations by year
2026: 23 · 2025: 25 · 2024: 1
FWCI
8.16
Type
journal article
Language
en
Open access
Licence
Locations
  • Empirical Software Engineering (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by)
  • White Rose Research Online (University of Leeds, The University of Sheffield, University of York) (submittedVersion, cc-by)
  • Open Repository and Bibliography (University of Luxembourg) (submittedVersion, cc-by)
  • Figshare (submittedVersion, cc-by-nc-sa)
  • University of Limerick Institutional Repository (University of Limerick) (submittedVersion, cc-by-nc-sa)
  • arXiv (Cornell University) (cc-by)
  • University of Limerick Institutional Repository (University of Limerick)
Funders
  • Fonds National de la Recherche Luxembourg (C22/IS/17373407/LOGODOR)
  • Natural Sciences and Engineering Research Council of Canada
Topics
  • Physical Sciences › Computer Science › Computer Networks and Communications › Software System Performance and Reliability
  • Physical Sciences › Computer Science › Information Systems › Software Engineering Research
  • Physical Sciences › Computer Science › Software › Software Testing and Debugging Techniques
Keywords
log parsing, log anomaly detection, deep learning, distinguishability

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Impact of log parsing on deep learning-based anomaly detection. https://app.cochat.ai/papers/10.1007/s10664-024-10533-w
The paper
Khan, Z. A., Shin, D., Bianculli, D., & Briand, L. C. (2024). Impact of log parsing on deep learning-based anomaly detection. Empirical Software Engineering. https://doi.org/10.1007/s10664-024-10533-w

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.