Project Title: Federated Learning for Privacy-preserving Video Caching Network
Funding Agency: National Science Foundation (USA), Institute of Information and Communication Technology Planning and Evaluation (Korea)
Award Number: 2152646 (USC)

Project summary:

Caching at the wireless edge is a promising method for improving the spectral efficiency and quality of video streaming. Caching strategies usually require knowledge of the video demands, as well as the historical preferences of the individual users. This information needs to be combined with information about the wireless network structure and state to develop deterministic or stochastic caching policies. Yet, in many cases it is undesirable to share detailed user proles with the cache and/or wireless network operator. The main motivation for avoiding such sharing is user privacy; furthermore video streaming providers do not want to share user popularity since this is important condential business information. A promising way out of this dilemma is federated learning (FL), which allows localized training and exchange of machine learning models. The current project thus investigates the advanced federated learning technique for an intelligent and privacy-preserving video caching and computing system and to novel federated learning techniques that preserve user privacy while retaining eciency of caching.

Specifically, the goals are to:

  1. Develop techniques that do not require transmission of models to a central server, thus further strengthening privacy
  2. Develop algorithms for hybrid global-local models that take account of the fact that video popularity is a global descriptor while also showing some local (spatio-temporal) variations
  3. Incorporate heterogeneous input data at UE and servers
  4. Directly learn caching strategies without taking the detour via separate learning of video popularity and wireless network state, thus improving both efficiency and privacy
  5. Develop an integrated system incorporating all of these aspects