Covertness-Oriented Water-to-Air Optical Wireless Semantic Communications

Abstract

Optical wireless communication (OWC) enables high-rate and low-latency direct data transmission between water and air, whereas wave refraction and particle scattering give rise to the issue of information leakage. To improve the energy efficiency and security in transmission, we propose a covertness-oriented semantic communication scheme for water-to-air (W2A) OWC systems in this paper. Based on a theoretical derivation regarding the monotonic correlation between the sparsity ratio of transmission encoding vectors and covert performance metrics, we introduce a sparsity-regularized semantic transmission scheme that extracts task-relevant semantic features, which significantly reduces the transmitted data volume and thereby lowers the probability of being detected by adversarial listeners. Aimed at underwater monitoring tasks, we establish a dedicated dataset with comprehensive coverage of core underwater anomalous object categories: the underwater anomaly monitoring (UAM) dataset. Based on this dataset, we perform thorough training and optimization on the parameters of the proposed scheme and test the scheme effectiveness. Simulation results demonstrate that the proposed scheme achieves high accuracy in anomaly detection under various noise conditions, while significantly improving signal covertness via adaptive sparsification. The related code and dataset will be available at https://github.com/qingqinghu-Ricky/Covert-semantic-communication-for-W2A-OWC upon publication.