Has F5 really been broken?

The publicly-available F5 software (F5Software) implementation takes a possibly compressed cover image, decompresses it if necessary, and embeds the hidden message during a second compression process. This procedure introduces a risk that the stego image goes through ‘double compression’. While this is not a problem from the embedding and extraction point of view, any steganalysis process trained on such a scheme will potentially detect artifacts caused either by the embedding process or the second compression process. In this paper we review published steganalysis techniques on F5. By re-implementing an isolated F5 embedding algorithm excluding the decompression and recompression process (F5Py), we show that published steganalysis techniques are unable to defeat F5 when its ideal operational condition is not violated. In other words, published techniques most likely detected the compression artifacts rather than the embedding process when the message size is not exceeding the optimum F5 capacity. This is an important fact that has been ignored before. Furthermore, we look for the optimum embedding rate for F5 in order for it to take advantage of matrix encoding for better embedding efficiency. From here we found that the low embedding rate considered for F5 in the previous works are actually relatively high for it. This is also important since bigger message size might degrade F5 to F4. In addition, we also verify that, as expected, steganalysis performance depends on the message size