Soft error decoding is used to decode steganographic messages embedded in image and audio signals. Content signals with auxiliary embedded data are analyzed to make determinations about changes made to the signals to embed the data. These determinations are weighted based on the confidence that they are correct and input to an error correcting process. In this manner, error correction of the auxiliary data takes advantage of soft error weighting to more effectively recover the steganographic message from content signals that undergo distortion and/or loss transformations after being embedded with steganographic data.