Every reviewer that I've met does plagiarism detection on their own. My advisor has even found instances where the author copy+pasted articles from the likes of the Wikipedia article on the subject or 'about.com'; such papers get desk-rejected without review, and in some cases, the author is forbidden from submitting to IEEE conferences for the rest of their career. That's how important plagiarism is.
Just copy+pasting the abstract into Google is often enough to find it. At least for computer science conferences, the conference organizers run the papers through additional plagiarism screening tools before publishing, but it's mostly the reviewers' job.
I don't think it's a bad idea, but it never occurred to me to do this. If you're familiar with the field in which you are reviewing papers, you'll know if something is not new, in which case it is rejected, quite aside from the possibility of the author having lifted some actual text.
I haven't seen this before, so maybe that is getting more common these years now.
Search on Google scholar: http://scholar.google.com/scholar?hl=en&as_sdt=1,5&a...
Do I really have to write IEEE for all violations? http://ieeexplore.ieee.org/search/searchresult.jsp?searchWit...
It's just searching for certain phrases in the text. In this case the first one I tried "This process is defined as binding process. In The immune system" was already enough!
And it's sad.