I think what they mean is that if your decompressed binary has long chunks of zeroes which compress really well, when it is encrypted (thanks DRM) those long series of zeroes get changed to non-reapeating data and are no longer easily compressible.

Usually: sizeof(compress(data)) <= sizeof(compress(encrypt(data)))

Why not first compressing then encrypting?

I suppose you might conceivably actually call them "contagious" zeroes rather than contiguous zeroes in that case.

You can't really compress encrypted data. If it's encrypted it looks more or less like random data. So any encryption is very likely to result in a (minor) size increase, when compared to the original.

That's what I was trying to say in my comment.