Keywords: heterogeneous distributed systems, reconstructive modeling, implicit anomalies, censored data, multivariate time series
UDC 004.9
DOI: 10.26102/2310-6018/2026.62.11.007
The relevance of the research is driven by the problem of anomaly detection in heterogeneous distributed systems under conditions of incomplete event logs and the absence of labeled data. According to reliability standards (GOST 27.002-2015, IEC 60050-191, IEC 60300-3-11, ISO/TR 12489), a hidden (latent) failure is defined as a state undetected by standard control and diagnostic methods during normal operation. In modern hardware-software complexes, such standard means include sensors and control system logs. However, in practice, a specific class of degradations arises: equipment undergoes physical degradation captured by multidimensional sensor time series yet entirely absent from the system's event logs. Since this scenario falls outside the classical definition of a hidden failure, this article aims to bridge the terminological gap by introducing the novel concept of an "implicit anomaly". A reconstructive modeling method for multidimensional time series is proposed, based on an LSTM autoencoder with subsequent verification through deviation density analysis within a time window. The problem of missing events is interpreted through the lens of censored data theory. Experimental validation on synthetic multidimensional time series demonstrated the method's ability to detect equipment degradation intervals unrecorded in system logs. The materials of the article are of practical value for developers of predictive analytics and monitoring systems, as the proposed approach increases the completeness of incident detection under conditions of partial observability of the controlled system.
1. Shcherbakov M.V., Safonova E.V., Shcherbakova N.L., Kravets A.G. Health Index Construction Method for Distributed Cyber-Physical Systems. In: Frontiers of Industrial Cyber-Physics. Cham: Springer; 2026. P. 71–89. https://doi.org/10.1007/978-3-032-02717-7_6
2. Safonova E., Kravets A., Shcherbakov M., Kizim A., Al-Gunaid M., Echin A. Industrial Environmental Impact Assessment Method Based on Detection of Complex Anomalies in Time Series. Applied System Innovation. 2024;7(5):89. https://doi.org/10.3390/asi7050089
3. Kleinbaum D.G., Klein M. Survival Analysis: A Self-Learning Text. New York: Springer; 2012. 700 p.
4. Konstantinov A.V., Utkin L.V. Deep forest for survival analysis in the framework of multivariate learning. In: XXII National Conference on Artificial Intelligence with International Participation: Volume 3, 06–10 October 2025, Saint Petersburg, Russia. Saint Petersburg: Saint Petersburg Federal Research Center of the Russian Academy of Sciences; 2025. P. 95–102. (In Russ.).
5. Chandola V., Banerjee A., Kumar V. Anomaly Detection: A Survey. ACM Computing Surveys. 2009;41(3):15. https://doi.org/10.1145/1541880.1541882
6. Blázquez-García A., Conde A., Mori U., Lozano J.A. A Review on Outlier/Anomaly Detection in Time Series Data. ACM Computing Surveys. 2021;54(3):56. https://doi.org/10.1145/3444690
7. Zhou Ch., Paffenroth R.C. Anomaly Detection with Robust Deep Autoencoders. In: Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2017), 13–17 August 2017, Halifax, NS, Canada. New York: ACM; 2017. P. 665–674. https://doi.org/10.1145/3097983.3098052
8. Zong B., Song Q., Min M.R., et al. Deep Autoencoding Gaussian Mixture Model for Unsupervised Anomaly Detection. In: Proceedings of the 6th International Conference on Learning Representations (ICLR 2018), 30 April – 03 May 2018, Vancouver, BC, Canada. 2018. URL: https://iclr.cc/virtual/2018/poster/126
9. Liu F.T., Ting K.M., Zhou Zh.-H. Isolation Forest. In: 2008 8th IEEE International Conference on Data Mining (ICDM 2008), 15–19 December 2008, Pisa, Italy. IEEE; 2008. P. 413–422. https://doi.org/10.1109/ICDM.2008.17
10. Hundman K., Constantinou V., Laporte Ch., Colwell I., Soderstrom T. Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding. In: Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2018), 19–23 August 2018, London, UK. New York: ACM; 2018. P. 387–395. https://doi.org/10.1145/3219819.3219845
11. Li D., Chen D., Jin B., Shi L., Goh J., Ng S.-K. MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks. In: Artificial Neural Networks and Machine Learning – ICANN 2019: Text and Time Series: 28th International Conference on Artificial Neural Networks: Part IV, 17–19 September 2019, Munich, Germany. Cham: Springer; 2019. P. 703–716. https://doi.org/10.1007/978-3-030-30490-4_56
12. Ruff L., Görnitz N., Deecke L., et al. Deep One-Class Classification. In: Proceedings of the 35th International Conference on Machine Learning (ICML 2018), 10–15 July 2018, Stockholm, Sweden. PMLR; 2018. P. 4390–4399.
13. Xu J., Wu H., Wang J., Long M. Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy. arXiv. URL: https://arxiv.org/abs/2110.02642 [Accessed 16th August 2026].
14. Tuli Sh., Casale G., Jennings N.R. TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data. Proceedings of the VLDB Endowment. 2022;15(6):1201–1214. https://doi.org/10.14778/3514061.3514067
Keywords: heterogeneous distributed systems, reconstructive modeling, implicit anomalies, censored data, multivariate time series
For citation: Safonova E.V., Alieva N.D., Shcherbakov M.V. Reconstructive modeling of multivariate time series for detecting implicit anomalies in censored data. Modeling, Optimization and Information Technology. 2026;14(11). URL: https://moitvivt.ru/ru/journal/article?id=2617 DOI: 10.26102/2310-6018/2026.62.11.007 (In Russ).
© Safonova E.V., Alieva N.D., Shcherbakov M.V. Статья опубликована на условиях лицензии Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NS 4.0)Received 02.09.2026
Revised 30.10.2026
Accepted 11.11.2026