MA: Federated learning of anomaly detection models for cloud monitoring
- © Copyright??
Topic Area - Federated learning of anomaly detection models for cloud monitoring
This topic area includes a larger set of different thesis topics
- Bachelor/Master: Implement and evaluate federated learning procedures (merging models) for ARIMA, BIRCH, HSTrees, or Deeplearning4j models
- Master: Combine active learning with federated learning
- Master: Combine federated learning with data privacy filters like Arx
All topics have to evaluate their approaches by detecting anomalies on cloud monitoring datasets, which are provided as metric data (CPU, memory, disc, etc. usage of cloud services) to you. The anomaly detection algorithms learn in an unsupervised procedure the normal behavior of the monitoring data stream of the individual cloud services. Through federated learning, the models should synchronize the learned behavior to gain an advantage from a broad set of services.
Prerequisites for working on this topic are advanced knowledge in Docker, algorithmic design, and very good programming skills in either Java or Python.