Joint Optimization for Secure and Reliable Communications in Finite Blocklength Regime

TL;DR

This paper develops a joint optimization framework for secure, reliable downlink communications in the finite blocklength regime, balancing secrecy rate, error probability, and information leakage.

Contribution

It introduces a novel alternating optimization approach with KKT-based solutions for secure precoding and error minimization in finite blocklength communications.

Findings

Proposed algorithm effectively balances security and reliability.

Simulation results validate the optimization approach.

Achieves improved secrecy and error performance in finite blocklength regime.

Abstract

To realize ultra-reliable low latency communications with high spectral efficiency and security, we investigate a joint optimization problem for downlink communications with multiple users and eavesdroppers in the finite blocklength (FBL) regime. We formulate a multi-objective optimization problem to maximize a sum secrecy rate by developing a secure precoder and to minimize a maximum error probability and information leakage rate. The main challenges arise from the complicated multi-objective problem, non-tractable back-off factors from the FBL assumption, non-convexity and non-smoothness of the secrecy rate, and the intertwined optimization variables. To address these challenges, we adopt an alternating optimization approach by decomposing the problem into two phases: secure precoding design, and maximum error probability and information leakage rate minimization. In the first phase, we obtain a lower bound of the secrecy rate and derive a first-order Karush-Kuhn-Tucker (KKT) condition to identify local optimal solutions with respect to the precoders. Interpreting the condition as a generalized eigenvalue problem, we solve the problem by using a power iteration-based method. In the second phase, we adopt a weighted-sum approach and derive KKT conditions in terms of the error probabilities and leakage rates for given precoders. Simulations validate the proposed algorithm.