Ounce of prevention

At the risk of this starting to look like a blog about comment spam, I have some additional thoughts on the matter.

I’ve made some changes to my comment forms here. The first is that the CGI script that comments get posted to is no longer the default mt-comments.cgi. I’ve created a clone of the comments script and renamed it fbda07e9fd3bb656bbf62c5b0ed6480e.cgi. That should stop bots that search for copies of mt-comments.cgi.

The next thing I’ve done is included a hidden field in each comment form that contains a MD5 hash of the entry ID and a secret word. Then I modified MT to check for that field. The comments script now creates a hash of the entry id and secret word and compares it to the one submitted with the comment. If that field isn’t submitted or it doesn’t match, the comment is rejected and the user is shown an error message.

But I wonder if these steps are useful at all. What I question is how spam bots are finding entries on which to comment. The entries that get the most spam comments here are those that have a large number of incoming links. The SimpleComments page is one of the hardest hit. That seems to suggest that bots are crawling from blog to blog, following links and posting comments.

This means that in order to post a comment, the bots must be parsing the HTML in order to find out if there’s a comment form on it. They aren’t apparently searching Google for common comment scripts otherwise the top search results would have the most spam comments.

Since the bots are parsing the HTML adding hidden form fields probably won’t deter them. If the authors of the bots have any brains whatsoever, they’re submitting all the hidden fields along with the forms. My hidden hash will be submitted by a bot just like it would by a person. What will probably be the biggest help is the thing that was easiest to do: changing the comment script name.

What else would be effective is changing the names of all the form fields. Making them short random strings would make it impossible for a bot to recognize the comment form using only the field names. People would be able to understand the form because of the labels, but bots would have to implement a large amount of fuzzy logic in order to recognize that “Name,” “Your Name:,” and other forms are really the same thing.