Decentralized Identity and Access Management for Distributed Machine Learning Systems

May 12th, 2022

Thesis Type
  • Master
Stefan Decker
Yongli Mou

Artificial Intelligence (AI) has more and more impact on people’s lives. According to the survey on the AI market, the global artificial intelligence software market has significantly increased in recent years. However, data privacy and security have become the main problems today’ AI is facing and hindering the broad application of AI techniques in many domains, since current AI approaches rely on large amounts of centralized data. As a new variety of AI, Federated learning (FL) brings learning to the edge or directly on-device based on decentralized data sources.

Under this background, there is a high demand for federated machine learning systems enabling data and model exchange. The design of identities for all kinds of resources, such as user credentials, clients, data and algorithms, and access control to the above resources are the basic requirements of such systems. However, the traditional centralized identity and access control methods have the following problems. Firstly, users lose control of their resources once uploading them. They cannot determine who has permission to access the resources and the duration of the permission. Secondly, people are concerned about data privacy while using provided services.

Thus, it is important to use the decentralized identity (DID) and a novel access control method to realize authorization, authentication, and accounting functionalities.  In the scope of this thesis, you will focus on Decentralized Identity and Access Control Management and develop a Blockchain-based DIDAM system for the federated machine learning system.

If you are interested in this thesis, a related topic, or have additional questions, please do not hesitate to send a message to


Fundamental knowledge on blockchain technologies, applications, and frameworks.