Identity-Preserving Covert Communication With Generative Perturbation

As covert communication gains prominence in wireless security, recent studies have proposed embedding hidden messages into physical-layer signals via subtle perturbations. However, most existing works focus solely on symbol-level modifications and often neglect the impact on transmitter identity, particularly the preservation of radio frequency (RF) fingerprints that are critical for device authentication and stealth. In this paper, we propose an Identity-Preserving Covert Communication (IPCC) framework that enables covert message embedding through carefully designed phase perturbations applied to pilot signals. We first present a Random Phase Perturbation (RPP) method, where each covert symbol is represented by a sequence of Gaussian-distributed phase perturbations. This sequence-level perturbation design mimics natural noise and enhances stealth, while implicitly offering fingerprint preservation due to its statistical consistency. To further improve identity consistency, we propose an Identity-Aware Generative Perturbation (IAGP) approach based on a Transformer-based conditional generator. The generator produces structured perturbations conditioned on both the covert message and the transmitter identity, enabling fine-grained control over both symbol expressiveness and fingerprint stability. Extensive experiments demonstrate that the proposed framework achieves high covert symbol classification accuracy with minimal impact on RF fingerprint recognition. This work provides a new perspective for covert communication by bridging statistical stealth and identity preservation through principled perturbation design.