betaFeedback

WorksEngineering

Physical Layer Security for Digital Semantic Communication: A Bit-Level Perspective

Abstract

Digital semantic communication requires reliable recovery of task-relevant information at a legitimate receiver while limiting source reconstruction by an eavesdropper. Uniform bit protection, however, does not account for the heterogeneous reliability and secrecy requirements of semantic bits. This paper proposes a bit-level physical layer security (PLS) framework that learns these requirements and incorporates them into physical-layer resource allocation. We first introduce a wiretap semantic feature channel whose two-stage training learns per-bit target bit-flip probabilities for the legitimate receiver (Bob) and the eavesdropper (Eve). These targets characterize the reliability requirements at Bob and the secrecy protection requirements against Eve. Based on the learned targets, we formulate a joint power and artificial-noise (AN) allocation problem for a multiple-input single-output wiretap channel under a worst-case angular-separation model. For a given target profile, the optimal information-AN power split admits a closed-form solution, and the per-bit power allocation is obtained by water-filling. We further develop a leakage-aware selection method among multiple learned-target profiles and extend the design to multiple-input multiple-output channels. Simulation results demonstrate that the proposed framework achieves a better trade-off between preserving reconstruction quality at Bob and degrading that at Eve than uniform power allocation and the evaluated learning-based secure semantic communication schemes.

Paper details

vlnc:W_963kmyyx
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
arXiv nonexclusive license
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Engineering

Not recorded for this paper

View statement
Looking for related papers…