Técnica de Aprendizagem não Supervisionada para Detecção de Falhas em Computação em Nuvem
Resumo
A crescente complexidade dos ambientes de computação em nuvem tem intensificado a necessidade de técnicas eficientes para detecção automática de falhas. Este artigo propõe uma abordagem baseada em aprendizado não supervisionado para análise de logs, utilizando Skip-Gram para geração de embeddings e K-Means para identificação de padrões anômalos. Os experimentos foram conduzidos em um ambiente OpenStack com injeção controlada de falhas dos tipos resizing e mismatch. Os resultados demonstraram AUC de 0,89 e acurácia de 87% para a falha resizing, e AUC de 0,857 e acurácia de 83% para a falha mismatch. Destaca-se o Recall de 1,0 para a falha mismatch, indicando detecção de 100% das falhas reais.Referências
Duarte, G. S. and Sousa, E. T. G. d. (2024). Cimut: Ferramenta de injeção de falhas em ambientes de nuvens por mutação. Trabalho de Conclus˜ao de Curso (Graduação). Estudo experimental realizado na plataforma OpenStack.
He, Z., Wang, R., and Liu, F. (2024). Logllm: Leveraging large language models for log analysis and anomaly detection. arXiv preprint arXiv:2402.01234.
Kim, D., Park, J., and Lee, H. (2025). Loggpt: Enhancing log anomaly detection using generative pre-trained transformers. In Proceedings of the ACM Symposium on Cloud Computing, pages 95–107.
Leshem, I. (2020). Skip-gram word2vec algorithm explained. [link]. Medium. Acesso em: 13 mar. 2026.
Li, H., Chen, R., and Wang, T. (2025). Graph-based anomaly detection for cloud systems using structural embeddings. In Proceedings of the IEEE International Conference on Cloud Computing, pages 312–319.
Liu, P., Zhao, K., and Sun, W. (2025). Intelligent failure detection in cloud systems using deep learning and telemetry analysis. In IEEE International Conference on Cloud Computing, pages 210–217.
Mohammed, A. K. et al. (2025a). Predictive failure detection in cloud infrastructure using multivariate telemetry log analysis with temporal convolution and attention-based deep learning. In International Conference on Cyber Resilience (ICCR), pages 1–8.
Mohammed, E. E., Naji, R. Y. S., Hussein, A. A., Saeed, M. A., and Al Selwi, R. A. M. (2025b). Anomaly detection system for secure cloud computing environment using machine learning. In 2025 5th International Conference on Emerging Smart Technologies and Applications (eSmarTA), pages 1–9.
Rana, O. F., Dustdar, S., et al. (2024). Multi-cloud computing: Architecture, challenges, and applications. IEEE Cloud Computing, 11(1):50–60.
Wang, H., Li, J., and Zhang, Y. (2024). A survey on aiops: Techniques, applications, and challenges. ACM Computing Surveys, 56(3):1–36.
Wu, J., Zhang, K., and Liu, Y. (2024). Graph neural networks for anomaly detection in distributed systems: A survey. IEEE Access, 12:33456–33472.
Zhang, T., Xu, H., and Li, C. (2024). Transformer-based log anomaly detection with self-supervised learning. IEEE Transactions on Network and Service Management, 21(2):789–802.
Zhou, L., Chen, M., and Wu, Q. (2024). Cloud-native systems: Architecture, challenges, and future trends. Future Generation Computer Systems, 150:45–60.
He, Z., Wang, R., and Liu, F. (2024). Logllm: Leveraging large language models for log analysis and anomaly detection. arXiv preprint arXiv:2402.01234.
Kim, D., Park, J., and Lee, H. (2025). Loggpt: Enhancing log anomaly detection using generative pre-trained transformers. In Proceedings of the ACM Symposium on Cloud Computing, pages 95–107.
Leshem, I. (2020). Skip-gram word2vec algorithm explained. [link]. Medium. Acesso em: 13 mar. 2026.
Li, H., Chen, R., and Wang, T. (2025). Graph-based anomaly detection for cloud systems using structural embeddings. In Proceedings of the IEEE International Conference on Cloud Computing, pages 312–319.
Liu, P., Zhao, K., and Sun, W. (2025). Intelligent failure detection in cloud systems using deep learning and telemetry analysis. In IEEE International Conference on Cloud Computing, pages 210–217.
Mohammed, A. K. et al. (2025a). Predictive failure detection in cloud infrastructure using multivariate telemetry log analysis with temporal convolution and attention-based deep learning. In International Conference on Cyber Resilience (ICCR), pages 1–8.
Mohammed, E. E., Naji, R. Y. S., Hussein, A. A., Saeed, M. A., and Al Selwi, R. A. M. (2025b). Anomaly detection system for secure cloud computing environment using machine learning. In 2025 5th International Conference on Emerging Smart Technologies and Applications (eSmarTA), pages 1–9.
Rana, O. F., Dustdar, S., et al. (2024). Multi-cloud computing: Architecture, challenges, and applications. IEEE Cloud Computing, 11(1):50–60.
Wang, H., Li, J., and Zhang, Y. (2024). A survey on aiops: Techniques, applications, and challenges. ACM Computing Surveys, 56(3):1–36.
Wu, J., Zhang, K., and Liu, Y. (2024). Graph neural networks for anomaly detection in distributed systems: A survey. IEEE Access, 12:33456–33472.
Zhang, T., Xu, H., and Li, C. (2024). Transformer-based log anomaly detection with self-supervised learning. IEEE Transactions on Network and Service Management, 21(2):789–802.
Zhou, L., Chen, M., and Wu, Q. (2024). Cloud-native systems: Architecture, challenges, and future trends. Future Generation Computer Systems, 150:45–60.
Publicado
25/05/2026
Como Citar
NUNES E SILVA, Carlos Manoel; SOUSA, Erica Teixeira Gomes de.
Técnica de Aprendizagem não Supervisionada para Detecção de Falhas em Computação em Nuvem. In: WORKSHOP DE TESTES E TOLERÂNCIA A FALHAS (WTF), 27. , 2026, Praia do Forte/BA.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 191-204.
ISSN 2595-2684.
DOI: https://doi.org/10.5753/wtf.2026.22965.
