[Paper Review] Fundamental Limits of Device-to-Device Private Caching with a Trusted Server under Uncoded Cache Placement and User Collusion

Kai Wan, Hua Sun+3 more|arXiv (Cornell University)|Dec 19, 2019

Caching and Content Delivery36 references4 citations

TL;DR

This paper proposes a novel D2D caching scheme with a trusted server to achieve information-theoretic privacy of user demands, even under user collusion, by using the server to coordinate coded multicast transmissions. It establishes achievable schemes and tight converse bounds, showing the scheme is within a constant factor of the optimal performance for private caching with uncoded placement.

ABSTRACT

In the coded caching problem, as originally formulated by Maddah-Ali and Niesen, a server communicates via a noiseless broadcast link to multiple users that have local storage capability. In order for a user to decode the desired file from the coded multicast transmission, the demands of all the users must be globally known, which may violate the privacy of the users. To overcome this privacy problem, Wan and Caire recently proposed several schemes that attain coded multicasting gain while simultaneously guarantee information theoretic privacy of the users' demands. In Device-to-Device (D2D) networks, the demand privacy problem is further exacerbated by the fact that each user is also a transmitter, which should know the files demanded by the remaining users in order to form its coded multicast transmissions. This paper solves this seemingly infeasible problem with the aid of a trusted server. Specifically, during the delivery phase, the trusted server collects the users' demands and sends a query to each user, who then broadcasts coded multicast packets according to this query. The main contribution of this paper is the development of novel achievable schemes and converse bounds for D2D private caching with a trusted server, where users may be colluding (i.e., when some users share cached content and demanded file indices), that are to within a constant factor of one another.

Motivation & Objective

• Address the challenge of demand privacy in Device-to-Device (D2D) caching networks, where users act as both transmitters and receivers.
• Overcome the inherent privacy leakage in D2D networks, where users must know others' demands to form coded transmissions.
• Ensure privacy even when colluding users share cached content and demand indices.
• Develop a scheme that leverages a trusted server to collect demands and coordinate secure, private coded multicasting.
• Establish fundamental limits through achievable schemes and converse bounds, proving near-optimality within a constant factor.

Proposed method

• Introduce a trusted server that collects all users' file demands during the delivery phase.
• Use the server to generate and distribute queries to each user, enabling them to compute and broadcast coded multicast packets without learning other users' demands.
• Design a caching scheme with uncoded placement to simplify implementation while preserving privacy.
• Formulate a novel transmission strategy where each user broadcasts coded packets based only on its own demand and the server-provided query.
• Derive converse bounds to characterize the fundamental limits of private caching under user collusion.
• Analyze the system under the assumption of global knowledge of user demands at the server, ensuring privacy at the user level.

Experimental results

Research questions

• RQ1 What is the fundamental limit of private caching in D2D networks when users may collude and use uncoded placement?
• RQ2 Can a trusted server enable private coded multicasting in D2D networks without exposing user demands to individual users?
• RQ3 How close can achievable schemes come to the theoretical limits of private caching under user collusion?
• RQ4 What is the impact of user collusion on the private caching gain in D2D networks?
• RQ5 Can a constant-factor approximation of the optimal private caching rate be achieved with a trusted server?

Key findings

• The proposed scheme achieves private coded multicasting in D2D networks with a trusted server, ensuring that no user learns the demands of others.
• The scheme is robust under user collusion, where colluding users cannot infer each other's demands due to the server-mediated query distribution.
• The authors establish both achievable schemes and converse bounds, showing that the performance is within a constant factor of the optimal rate.
• The fundamental limits of private caching with uncoded placement are characterized under the assumption of a trusted server.
• The scheme maintains the coded multicasting gain while satisfying information-theoretic privacy, even in the presence of colluding users.
• The analysis confirms that the trusted server enables a practical and provably optimal solution to the private D2D caching problem.