CS/Network

[논문 리뷰] Mobile Traffic Forecasting for Network Slices: A Federated-Learning Approach

동현 유 2023. 5. 18. 16:05

Authors:

H. P. Phyu, D. Naboulsi and R. Stanica, 

 

Journal/Conference:

2022 IEEE 33rd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Kyoto, Japan, 2022, pp. 745-751

 

Source:

https://ieeexplore.ieee.org/document/9977882

 

Presentation meterials:

Mobile Traffic Forecasting for Network Slices, A Federated-Learning Approach.pdf
1.29MB

 

 

 

Abstract

Network slicing is one of the cornerstones for next-generation mobile communication systems. Specifically, it enables Mobile Virtual Network Operators (MVNOs) to offer various types of services over the same physical infrastructure owned by an Infrastructure Provider (InP). To satisfy the dynamic user requirements and ensure resource efficiency, MVNOs need to estimate the future traffic demand in advance, to pre-allocate/reconfigure the resources at the base stations. However, this per-slice traffic forecasting exploits information that is clearly sensitive for the MVNOs from a business point of view, and which might even disclose private data regarding some users. Hence, it is vital for MVNOs to ensure data privacy while conducting traffic forecasting. Bearing this in mind, we propose the Federated Proximal Long Short-Term Memory (FPLSTM) framework, which allows MVNOs to train their local models with their private dataset at each base station without compromising data privacy. Simultaneously, an InP global model is updated through the aggregation of local models weights. Prediction results obtained by training the models on a real-world dataset indicate that the forecasting performance of FPLSTM is as accurate as state-of-the-art solutions, while ensuring data privacy, computation and communication cost efficiency.

 

Proposed method

  • FPLSTM is an efficient privacy-preserving multi-slice traffic forecasting framework, both communication and computation-wise, by sharing only parameters of the models.
  • very straightforward approach.

Discussion

  • There may be scaliability issues because of the network communication cost.
  • To check the feasibility of deployment in the real network, additional detailed figures should be provided about the size of LSTM's parameters versus the size of real dataset being transferred.
  • Since MVNOs may manage tens of hundreds gNBs for their services, so the limitation about scaliability should be experimented and clearly delivered.