Decentralized Federated Learning: A Comprehensive Survey and a New Blockchain-based Data Evaluation Scheme

Blockchain and Deep Learning (DL) are two of the most revolutionary concepts in the field of Computer Science. Both have made astounding leaps in research and application areas such as Finance, Healthcare, Internet of Things, and many more. Federated Learning (FL) is a type of distributed Deep Learning framework, in which the model is trained locally on each device and the trained gradients are sent to a central server which aggregates them and creates a global model. This helps ensure the data privacy of the user as the data never leaves the local device. However, this dependency on the central server can lead to various issues such as lack of transparency and communication bottleneck. Making this process decentralized can help address these issues. In this review, a detailed survey on using blockchain in federated learning is presented. This review also focuses on how can we use blockchain to make federated learning more transparent and decentralized to protect the privacy of the user. We also discuss the major strengths and drawbacks of each approach and further present a few ideas of our own, regarding some of these challenges and ways on how can these be improved. A new scheme to evaluate data using miners as well as methods to reduce storage overhead in decentralized federated learning are discussed in this paper.