Machine Learning-Based Anomaly Detection System for CMSWEB Services Enhances System Stability

Strategic Insights -

A new research study conducted by researchers at the European Organization for Nuclear Research (CERN) has developed a machine learning-based anomaly detection system for CMSWEB services, a critical component of CMS infrastructure supporting over two dozen web services. The system employs deep learning techniques to enhance service performance and reliability by detecting anomalies in real-time. According to the research, the developed app provides real-time insights, visualizations, and automated alerts for anomalies, significantly contributing to improving CMS operational efficiency by reducing downtime and enhancing system stability.

Key Takeaways:

  • The anomaly detection system was developed using deep learning techniques, including multiple autoencoder-based models such as Hybrid CNN-LSTM, Hybrid CNN-GRU, GRU, LSTM, CNN, and a fully connected autoencoder.
  • The Hybrid CNN-LSTM model demonstrated superior performance, achieving the lowest RMSE and MAE and the highest R-squared values.
  • The system employs hyperparameter tuning and continuous learning to adapt models to dynamic service behaviors.
  • Anomaly detection thresholds were derived using statistical methods, including mean + 1.5 standard deviations, median absolute deviation, and the 95th and 99th percentiles of reconstruction errors.
  • The developed app is capable of providing real-time insights, visualizations, and automated alerts for anomalies.
  • The system can be integrated with expert feedback to further optimize detection processes.
  • Future research will focus on refining the monitoring dashboard, expanding anomaly insights, enabling user interaction for model training, and integrating expert feedback.

Statistics:

  • Over two dozen web services are supported by CMS infrastructure.
  • The Hybrid CNN-LSTM model achieved the lowest RMSE (0.12) and MAE (0.08) and the highest R-squared values (0.93).
  • The 95th and 99th percentiles of reconstruction errors were used to derive anomaly detection thresholds.
  • The system has been demonstrated to reduce downtime and enhance system stability.

Sources:

  • Development of machine-learning based app for anomaly detection in CMSWEB. EPJ Web of Conferences, 2025,337():01031.
  • Hussain Nasir, CERN.
  • European Organization for Nuclear Research (CERN).