Federated Learning-based Content Caching Strategy for Edge Computing

Abstract

In the realm of edge computing, effective content caching stands as a pivotal strategy to manage the exponential surge of mobile data within 5G networks. Content caching revolves around enabling the local storage of content in caches, ensuring swift and recurrent access to data. Yet, the challenge lies in accurately predicting the popularity of the files and thus requires caching due to constraints such as limited cache space and fluctuating file preferences. Conventional learning-based methods tackle this issue by gathering user data centrally for training purposes, to predict the file popularity. However, there arises a concern regarding user reluctance to entrust their private data to a central server. To address this, a novel solution is introduced in this paper called Federated Learning-based Content Caching (FLCC). FLCC operates by employing an enhanced Stacked Autoencoder in the edge devices without the necessity for centralized data collection during training. Utilizing a federated learning approach, FLCC prioritizes data privacy by aggregating user updates through federated averaging. This method implements a hybrid filtering mechanism based on a stacked autoencoder, training each user individually with their local data. The results from the FLCC approach showcase its superior cache efficiency when compared to traditional learning-based techniques. The proposed FLCC approach is a robust solution that upholds data privacy while enhancing content caching effectiveness in edge computing environments.