Audio Steganography using Convex Demixing

In the first part, we first introduce steganography (in chapter 1) not in the usual context of information security, but as a method to piggyback data on top of some content. We then focus on audio steganography and propose a new steganographic scheme in chapter 2 as well as a model for the noisy, analog communication channel we are considering (in section 3.2). The method we use is based on signal mixing, the science of constructing vectors in a way that their sum can then be split back into the original components. This is presented in chapter 4. The data recovery, or demixing, relies on convex optimization (presented in chapter 5) and some further signal processing detailed in chapter 6. In the second part we present our proof of concept implementation and show the results of simulation runs that have been made in order to study the properties of the overall system (chapter 7). We study how the different parameters of the system and the communication channel affect the performance of the steganographic scheme.Finally, we draw conclusions (chapter 8) about the findings and suggest (in chapter 9) what next steps could be taken in order to further study this steganographic scheme.