add post on mailman subscritpion spam

diff --git a/ralf/_posts/2018-06-02-mailman-subscription-spam.md b/ralf/_posts/2018-06-02-mailman-subscription-spam.md
new file mode 100644 (file)
index 0000000..237153c
--- /dev/null
+++ b/ralf/_posts/2018-06-02-mailman-subscription-spam.md
@@ -0,0 +1,59 @@
+---
+title: "Fighting Mailman Subscription Spam: The Easy Way"
+categories: sysadmin
+---
+
+I recently noticed that both of the Mailman setups that I am running are being
+abused for *subscription spam*: Bots would automatically attempt to subscribe
+foreign email addresses to public mailing lists, resulting in a subscription
+notification being sent to that address.  I am still extremely saddened by the
+fact that this is a thing---whoever sends this spam has no direct benefit and no
+way of selling anything (they don't control the content of the message); the
+only effect is to annoy the owner of that email address, the victim.  That seems
+to be enough for some. :(
+
+Oh, and my servers' reputation goes down because people mark these emails as
+spam.  So, more than enough reasons to try and stop this.
+
+<!-- MORE -->
+
+### The Big Guns
+
+My first reaction was to go and look for a way to add a CAPTCHA to the
+subscription page.  Unfortunately, Mailman itself does not support a CAPTCHA (at
+least not Mailman 2), and the existing patches I found were all about adding
+support for Google's reCAPTCHA.  I am not going to expose my users to Google's
+tracking like that, nor am I willing to actively discriminate against people not
+having Google accounts (reCAPTCHA is much more annoying if Google can't track
+you because you are not logged in), so reCAPTCHA was clearly not an option.
+Instead, the plan was to look at one of these patches and implement a simple
+question-and-answer CAPTCHA myself.
+
+### Keep It Simple
+
+But then, while just getting started on this and browsing the Mailman sources, I
+found out about `SUBSCRIBE_FORM_SECRET`.  `SUBSCRIBE_FORM_SECRET` is a Mailman
+config option that, once set to a random string, will make Mailman embed a
+[CSRF token](https://www.owasp.org/index.php/Cross-Site_Request_Forgery_(CSRF)_Prevention_Cheat_Sheet)
+into the subscription form.  Mailman will also enforce that the form must be
+submitted *at least* five seconds after it was generated.  Since the bots that
+have found my servers so far are much less patient than that, just setting
+`SUBSCRIBE_FORM_SECRET` was enough to completely get rid of the subscription
+spam.
+
+So, if you are reading this and running a Mailman installation: **Please set
+`SUBSCRIBE_FORM_SECRET` and protect your setup against abuse!**  Just run `pwgen
+16` to get some random string, and then add `SUBSCRIBE_FORM_SECRET = "<random
+string here>"` to `/etc/mailman/mm_cfg.py`.  It's really that simple!  Just a
+[four-line patch in my Ansible playbook](https://git.ralfj.de/ansible.git/commitdiff/937b170594be82e500ae726dc47de8ca9ef3dfcf)
+to get this rolled out to all servers.  Note that you need to be at least on
+Mailman 2.1.16 for this to work; all currently supported versions of Debian come
+with a recent enough version (if you use backports on Debian 7 "Wheezy").
+
+The more people do this, the more it will help to stop this kind of spam.  Or
+rather, it'll force the spammers to upgrade their game.  I assume eventually I
+*will* have to add a CAPTCHA.  Or maybe there is a simple and reliable way to
+migrate to Mailman 3 before that happens---and maybe that will have a CAPTCHA.
+(Though, from a quick search, it doesn't seem like it does, which I find pretty
+surprising.  If my tiny servers are abused like this, I assume it's a really
+common problem and Mailman should protect against it per default.)