As part of my exchange with the University of Montreal , I wrote, in collaboration with Pierre Turlan, a paper on subliminal channels (supervised by Stefan Wolf). Most of my work was focused on the DSA algorithm. The project includes a practical implemenation of subliminal channel in DSA.
The main idea of the project is to produce software capable of producing perfectly valid DSA signatures although in fact each signature reveals the signer's private key! This illustrates the necessitated use to open source cryptographic software because it makes no sense to trust a software simply because it seems to work correctly. It is necessary that the sources are checked to prevent mistakes as well as fraud.
For more information on this project, as well as the original paper in PDF, please see (in French):