Joint Information Theoretic Secrecy and Covert Communication in the Presence of an Untrusted User and Warden

TL;DR

This paper explores joint secrecy and covert communication strategies in a SIMO system with an untrusted user and warden, optimizing transmission to maximize rate under security and covertness constraints.

Contribution

It introduces a novel joint secrecy and covert communication framework considering imperfect CSI and untrusted nodes, with a tractable optimization approach.

Findings

Imperfect CSI at Carol impacts rate more than at Bob.

Successive convex approximation effectively solves the non-convex optimization.

Joint strategies improve security and covertness in SIMO systems.

Abstract

In this paper, we investigate joint information theoretic secrecy and covert communication in a single-input multi-output (SIMO) system where a transmitter (Alice) is communicating with two legitimate users (Bob and Carol). We consider that an untrusted user and a warden node are also present in the network attempting to attack the secure and covert communications to Bob and Carol, respectively. Specifically, Bob requires secure communications such that his messages from Alice are not decoded by the untrusted user, while Carol requires covert communications such that her messages from Alice are not detected by the warden. To do so, we consider that Alice transmits Carol's messages during selected time slots to hide them from the warden while also transmitting Bob's messages in each time slot contentiously. We formulate an optimization problem with the aim of maximizing the average rate subject to a covert communication requirement and a secure communications constraint. Since the proposed optimization problem is non-convex, we utilize successive convex approximation to obtain a tractable solution. We consider practical assumptions that Alice has imperfect knowledge of the warden's location and imperfect channel state information (CSI) of Bob and Carol. Our numerical examples highlight that the imperfect CSI at Carol has a more detrimental impact on the average rate compared to imperfect CSI at Bob.