Trac spam
51 views
Skip to first unread message
Aymeric Augustin
Mar 1, 2014, 7:07:25 AM3/1/14
to django-d...@googlegroups.com
Hello,
If you’ve subscribed to django-updates, you must have noticed that spam levels have increased badly over the last weeks.
The spam filters are fending off about a thousand spam attempts a day, but a small fraction (0,1-0,5%) gets through. Of course, we’re deleting it as soon as we see it in order to discourage spammers.
In order to improve the situation, I’ve updated the spam filter plugin to its latest version. I don’t know if that’s going to change much. I’ve also fed a lot of ham to the built-in bayesian filter, because it needs a fairly balanced sample of spam and ham to work well. Unfortunately the kind of spam that goes through looks designed to defeat bayesian filtering.
Finally, when there’s a popular spamming topic, we can reject it by adding banned regexes to the BadContent wiki page. For instance, https://code.djangoproject.com/wiki/BadContent?action=diff&version=76 would have blocked 42 spams since January 17th. (I discovered this feature this morning.) Triagers, when you delete spam, feel free to add to this list!
We’re still quite far from losing this fight, our problem is simply that:
- we want to allow anonymous bug reports
- no one really masters Trac’s spam filtering system.
--
Aymeric.
If you’ve subscribed to django-updates, you must have noticed that spam levels have increased badly over the last weeks.
The spam filters are fending off about a thousand spam attempts a day, but a small fraction (0,1-0,5%) gets through. Of course, we’re deleting it as soon as we see it in order to discourage spammers.
In order to improve the situation, I’ve updated the spam filter plugin to its latest version. I don’t know if that’s going to change much. I’ve also fed a lot of ham to the built-in bayesian filter, because it needs a fairly balanced sample of spam and ham to work well. Unfortunately the kind of spam that goes through looks designed to defeat bayesian filtering.
Finally, when there’s a popular spamming topic, we can reject it by adding banned regexes to the BadContent wiki page. For instance, https://code.djangoproject.com/wiki/BadContent?action=diff&version=76 would have blocked 42 spams since January 17th. (I discovered this feature this morning.) Triagers, when you delete spam, feel free to add to this list!
We’re still quite far from losing this fight, our problem is simply that:
- we want to allow anonymous bug reports
- no one really masters Trac’s spam filtering system.
--
Aymeric.
Reply all
Reply to author
Forward
0 new messages