Using key positions in steganography using DCT

Question

I have a question about steganography. Once the DCT matrix's are computed and the LSB's are replaced can we assume that the location at which the data is entered in the DCT of the cover image is know at the receiver? If not then how is the image reconstructed?

Answer 1

Generally speaking, the extraction method assumes the knowledge of the embedding rules so it can reverse them. Deterministic processes are hardcoded, while for stochastic processes you need to provide additional information to replicate their state.

For example:

• If you modify a specific DCT coefficient in an 8x8 block, the extraction will have to know that.

• If you hide a bit in either coefficient A, or B, based on some rule e.g, the lsb of some coefficient C, the extraction will do the same. That is, check what coefficient C says to decide whether to extract the data from coefficient A or B.

• If the coefficient, or order of coefficients within a block, is determined by a prng, both the embedding and extracting processes have to initialise that with the same seed. That takes the form of a password which the receiver must know in order to read the correct sequence of coefficients to extract the data.

And so on, with more complex rules.