In this paper, we tackle the privacy issue in cybertwin with federated learning (FL), which exploits clients to collaboratively train a machine learning (ML) model without the awareness of private data. We develop an estimation scheme to reveal the data distribution of local dataset residing in the clients without the awareness of private data. We consider two scenarios in FL: 1) the server could receive the individual trained model for each selected device; 2) the server could receive the aggregated model from the selected clients. We formulate two device selection problems for training performance improvement. We develop two online learning algorithms to tackle the selection problems for both individual model uploading and aggregated model uploading. The proposed algorithms are demonstrated, aiming to avoid privacy leakage and extra computation in the clients for 6G. We validate the effectiveness of the proposed client selection algorithms with sufficient experiments in cybertwin-driven 6G networks.
- 6G mobile communication,
- client selection,
- Computational modeling,
- Cybertwin-driven 6G,
- Data models,
- Data privacy,
- federated learning,
- imbalanced distribution,
- Performance evaluation,
- privacy-preserving,
- Servers,
- Training
IR Deposit conditions:
OA version (pathway a) Accepted version
No embargo
When accepted for publication, set statement to accompany deposit (see policy)
Must link to publisher version with DOI
Publisher copyright and source must be acknowledged