| Abstrakt: |
There are significant challenges in machine learning models due to information security and data privacy issues. In traditional machine learning approaches, the data used to train the centralized model may be sensitive leading to privacy issues. Federated learning overcomes these issues by following the concept of bringing the code to the data instead of data to the code. In today's era of digital transformation, researchers and industries are focusing on cloud-based technologies. New approaches such as microservice models, docker, kubernetes, containers, and virtual machines are being used by IT industries to develop and deploy applications. We make use of software containers to deploy federated learning models and then we identify security vulnerabilities and resource usage in federated learning model. Flask, docker, cAdvisor, prometheus, and grafana are integrated to monitor the resource usage of the federated learning model in terms of CPU utilization, memory utilization, network traffic usage, and disk usage. The generated federated learning model is deployed on docker, and security vulnerabilities and performance are analyzed using open-source tools. The result shows that the generated model is secure and performs efficiently. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink