Federated Learning Based Proactive Content Caching in Edge Computing

Abstract

Content caching is a promising approach in edge computing to cope with the explosive growth of mobile data on 5G networks, where contents are typically placed on local caches for fast and repetitive data access. Due to the capacity limit of caches, it is essential to predict the popularity of files and cache those popular ones. However, the fluctuated popularity of files makes the prediction a highly challenging task. To tackle this challenge, many recent works propose learning based approaches which gather the users' data centrally for training, but they bring a significant issue: users may not trust the central server and thus hesitate to upload their private data. In order to address this issue, we propose a Federated learning based Proactive Content Caching (FPCC) scheme, which does not require to gather users' data centrally for training. The FPCC is based on a hierarchical architecture in which the server aggregates the users' updates using federated averaging, and each user performs training on its local data using hybrid filtering on stacked autoencoders. The experimental results demonstrate that, without gathering user's private data, our scheme still outperforms other learning-based caching algorithms such as m--greedy and Thompson sampling in terms of cache efficiency.

Deep-Reinforcement-Learning-Based Mode Selection and Resource Allocation for Cellular V2X Communications

“…In fact, the training data is always distributed at vehicles and unlikely to be uploaded considering bandwidth overhead and privacy issues. Fortunately, federated learning has the potential to realize distributed learning [22], [23]. To achieve high cache efficiency as well as to protect users' privacy, Yu et al [22] propose a federated learning based proactive content caching scheme which does not require to gather users' data centrally for training.…”

Joint Content Placement and Storage Allocation Based on Federated Learning in F-RANs

“…User preferences can be predicted in advance or on a regular basis (e.g., hourly, daily, or weekly) through systematic learning and analysis of user social behavior [ 36 , 37 ]. In this paper, considering the privacy security of users, we adopt the federated learning method [ 7 ] to accurately predict the content popularity in the region.…”

“…Due to the limited storage capacity of edge nodes, the network performance can be effectively improved by predicting the content popularity and actively caching the most popular one. However, most existing caching schemes are designed for highly controlled environments where users need to upload local private data to a central server, which may pose privacy and security risks [ 7 ]. Furthermore, with the increase of the number of users and data, the unreliability and communication cost of wireless networks cannot be ignored.…”

A Classification of the Enabling Techniques for Low Latency and Reliable Communications in 5G and Beyond: AI-Enabled Edge Caching

“…The main reason for the poor performance in the LRU and LFU solutions is that, the algorithms in LRU and LFU do not consider the popularity of contents in the future. As a result, the solutions do not adapt well to the dynamically changing content popularity and they achieve low cache efficiency [89], [113]. For instance, the LFU framework is not able to reach a good performance in IoT environment because it does not consider the saltation and timeliness of the IoT data popularity [113].…”

“…Although many studies have been done to address the issues of edge caching and in-network caching, many issues remain open for more research. For example, the work in [34], [89], [113], [120], [131] considered the future popularity of content while making cache decisions. In [113], various fundamental questions about IoT data popularity and related popularity-based caching were considered.…”