Andreas Westfeld and Gritta Wolf [WW98] have made an interesting proposal for data hiding in a videoconferencing system.
Basically, the technique used is DCT coefficient manipulation, which becomes very valuable in the context of videoconferencing.
To achieve high frame rates on narrow band digital networks, videoconferencing applications compress every frame with a differential lossy compression, which means that only differencies between successive stills are first compressed and then broadcast.
The advantages in manipulating DCT coefficients in this situation is that the overall effect is very similar to the modification due to slight movements of the camera. This renders the embedding technique almost invisible, and the authors claim that one can differentiate between the original and the stego images, but can't tell which is which.
Moreover, one of the threat of detection is comparison between successive similar frames that would magnify the presence of encoding. Since videoconferencing systems only broadcast the differences between successive frames, then this threat is automatically voided.
Other attacks are very unlikely to succeed in detecting or extracting the stego-message, since the added noise is very similar to the original one, and the data are encrypted before embedding, making them more difficult to find.
The data rate for the described technique has been reported as up to 8KB/s (a GSM conversation) embedded into an ISDN videoconference.
This points it out as a valuable classical steganography scheme, due to the high data rate and good stealth.
[WW98]
Previous: other methods for still images
Next: other media
Matteo Fortini