Re: [Mailman-Developers] Remediation for fake member creation

----- Original Message -----
yes but protecting the web form from non-human subscription is a good step to take
WDOT?
Can't you send the email subscription request to moderation before the email confirmation is sent?
Not ideal, but it is kind of like emergency moderation.

Franck Martin writes:
Can't you send the email subscription request to moderation before the email confirmation is sent?
The option "subscription needs approval" is available, and I use it for my student lists, etc. They're closed lists initially populated with "mass subscribe", but students often want to use cellphone or webmail addresses in addition to or in preference to their university addresses. In general, if the moderator knows the users well, there's often no point in confirmation. Eg, in my case I've almost always received personal mail from the address (it's preferred, or at least frequently used) in question if the student is on my list, so I know it's theirs.
There is also an option "confirm and approve". I believe it means "confirm, *then* approve", and I think that's the right order. First, it prevents an attack on the moderator using faked addresses, and makes it a lot more expensive to attack the moderator with real addresses. I have seen such attacks on occasion for going on 25 years now; it's not a nightmare, it's a real problem.
Second, moderators are a scarce resource. In many cases the moderator will need to follow up out of band (for example, I recently subscribed to a closed list, and the moderator texted me on Telegram to make sure it was me). In that case, either way the "victim" has to deal with an additional contact -- we can't save them the effort, we can only reduce load for the moderator by asking the user to confirm first. Then if the user drops it on the floor, the moderator has no work to do. Of course there would be cases where the moderator would refuse the request before confirmation, but I think that would depend on the moderator knowing that there were attacks via her list. On balance, I strongly favor protecting the moderator here.
Finally, for open lists, which currently are configured confirm-only, I don't see how the moderators would have any idea whether it was a legitimate request or an attack, unless it was repeated to the same list -- and even then it would have to be a memorable address.
Bottom line: I see no reason to default "needs approval" on for Mailman as we distribute it, unless we discover that "moderator knows subscribers" is by far the most common case. cPanel might think otherwise for their user base, I don't know. But not the typical open source project or discussion list, which I believe is by far the majority of non-cPanel (etc) Mailman lists. And the option is always available to turn on if you realize your list is being abused that way.
Steve

Franck Martin writes:
Can't you send the email subscription request to moderation before the email confirmation is sent?
The option "subscription needs approval" is available, and I use it for my student lists, etc. They're closed lists initially populated with "mass subscribe", but students often want to use cellphone or webmail addresses in addition to or in preference to their university addresses. In general, if the moderator knows the users well, there's often no point in confirmation. Eg, in my case I've almost always received personal mail from the address (it's preferred, or at least frequently used) in question if the student is on my list, so I know it's theirs.
There is also an option "confirm and approve". I believe it means "confirm, *then* approve", and I think that's the right order. First, it prevents an attack on the moderator using faked addresses, and makes it a lot more expensive to attack the moderator with real addresses. I have seen such attacks on occasion for going on 25 years now; it's not a nightmare, it's a real problem.
Second, moderators are a scarce resource. In many cases the moderator will need to follow up out of band (for example, I recently subscribed to a closed list, and the moderator texted me on Telegram to make sure it was me). In that case, either way the "victim" has to deal with an additional contact -- we can't save them the effort, we can only reduce load for the moderator by asking the user to confirm first. Then if the user drops it on the floor, the moderator has no work to do. Of course there would be cases where the moderator would refuse the request before confirmation, but I think that would depend on the moderator knowing that there were attacks via her list. On balance, I strongly favor protecting the moderator here.
Finally, for open lists, which currently are configured confirm-only, I don't see how the moderators would have any idea whether it was a legitimate request or an attack, unless it was repeated to the same list -- and even then it would have to be a memorable address.
Bottom line: I see no reason to default "needs approval" on for Mailman as we distribute it, unless we discover that "moderator knows subscribers" is by far the most common case. cPanel might think otherwise for their user base, I don't know. But not the typical open source project or discussion list, which I believe is by far the majority of non-cPanel (etc) Mailman lists. And the option is always available to turn on if you realize your list is being abused that way.
Steve
participants (2)
-
Franck Martin
-
Stephen J. Turnbull