VS 2010 Comparing Published Files?

We publish official documentation into pdf's, tiffs, and cals depending on the document type. Recently, we have discovered that somewhere in our publishing process, the system can hiccup and cause a character or two to become missing. It's intermittent and not consistent. We have a fix for it, but unfortunately have hundreds of thousands already published in the repository. It was brought up that we may want to consider developing a batch comparison tool that would compare a newly published version to the original published version to identify the problem documents. There was multiple thoughts on how to do this:

1.) Compare file sizes... unfortunately, any variation will cause the sizes to vary and small variations may not accurately reflect in it.

2.) Binary comparison... not sure how this would work at a file level like this

3.) Hash MD5 Compare... not sure how this would work either.

So I thought I'd ask those out there with some similar experience, what is the best way using the .Net Framework to compare one tiff to another, one pdf to another, and one cal to another?