This paper presents several watermarking methods preventing the estimation of the secret key by an adversary. The constraints for secure embedding using distribution matching, where the decoding regions rely implicitly on the distribution of the host signal, are first formulated. In order to perform informed coding, different decoding regions are associated with the same message using an appropriate partitioning function. The minimization of the embedding distortion is afterwards casted into an optimal transport problem. Three new secure embeddings are presented and the performances of the proposed embedding functions regarding the AWGN channel for different WCRs are evaluated. Depending on the embedding and noise distortions, informed secure coding can outperform classical secure coding or classical insecure coding such as ISS or SCS.