Can Sonar be used to extract a sound?

Not likely - but depends.
If for example it's a stereo file comprised of two mono tracks where the originator simply took the backwards copy and put it on one side and the regular mono track on the other - sure. This is unlikely however.
The best analogy I can think of for what it sounds like you are trying to do is like "unpainting a painting" - for example you want to remove all the black from an oil painting. Or all the brush strokes the artist did last Monday. There are tools (solvents) that might do it, but not perfectly and not without some work.
No doubt there are audio tools that could do something analogous but not easily, and not perfectly. Not what SONAR is intended to do - unless the information you want is present as separate tracks in the original file already.

You might be able to use things like iZotopes RX4 or maybe Sonys Spectral Layers 3 but things like this are usually only undoable at the construction stage not once the audio has been rendered to its own audio file

Thanks for the responses everyone. I kinda figured that nothing would be possible with Sonar... or any software for that matter. :(

gswitz
I can't imagine a mathematical algorithm that could do this without first having a copy of the original sound file.

If you merge any two wave forms, you can reverse the process as long as you retain one of the two original forms.

This is the answer barring simple channel separation. Is cryptography, but the key would have to exist on both machines. Using another (easily accessible) song would work, since to use the encrypted song onto itself would require an original copy to be sent (sort of ruins the point if an unencrypted copy exists). The second machine should only need to reverse phase (null out) the original key.