Hi,
Reading back a few threads, I see that we've not yet been able to
successfully supply a method for users to report security problems,
privately. I think this is very important to the integrity of Habari,
especially as it gains more users.
I understand that we have some problems with respect to this and
Google Groups.
So, I've taken the liberty of setting up habaris...@iconoclast.caedmon.net
(my server)—it automatically rewrites the headers and sends mail to
the private list (as you probably saw with my tests). Michael (Harris)
helped me set up this first step. If this isn't cool, feel free to
remove permission for this address from the list/group.
Unfortunately, Google Groups hacks up the email and drops a bunch of
headers (or so it seems), so I propose this the following.
Set up a "Report a security issue" link on hp.o (and link it in trac,
if possible) that takes the user to a form that asks for:
Name (optional), Email (optional), Description of vulnerability. The
processor of this form is a script that collates the data and emails
it (in the body) to habaris...@iconoclast.caedmon.net, which
relays it to the private list.
This bypasses the groups problem, keeps the address non-public and
allows us to CC the reporter on replies (if we choose). As a policy, I
think we should mail all reporters to let them know we've received
their report.
Thoughts? (perhaps this should be a thread on habari-dev?)
S
If we're going to use a form instead of having users mail security@hp.o directly, why do we need an email relay script? We could easily just set the email's FROM header to be security@hp.o, which is already set as a member of the list.
We would still want security@hp.o to be available, but having an additional form on the site would probably be a good thing.
We would still want security@hp.o to be available, but having an additional form on the site would probably be a good thing.
Depends what MTA is installed, but most will allow us to do something
like this in /etc/aliases:
security: |/path/to/script
(the script has to be in the appropriate location and must be +x)
It's working, currently, on my test setup (postfix).
S
From what I can tell, this works with exim, but might need some setup
with the weird router and transport stuff that Exim does (and I
haven't run exim in years, so I really don't know).
Anyone else?
S
Done. (-:
We now have a functioning secu...@habariproject.net (not .org yet).
Can someone please redirect the MX records for hp.o to point at the
same IP as www ?
S
Hi,
Sorry this was unclear.
Please change the MX for hp.o to point at www.hp.o instead of google.
Thanks.
S
You can't just forward it, because then we'd have to accept emails from any address to -private (since we don't know the address it's coming from). We had to setup a script to "rewrite" or "redirect" them with a from address we knew so we could add that to the membership of -private.
Surely when you forward it, google apps changes the from address?On 15 Dec 2008, at 12:38, Chris Meller wrote:You can't just forward it, because then we'd have to accept emails from any address to -private (since we don't know the address it's coming from). We had to setup a script to "rewrite" or "redirect" them with a from address we knew so we could add that to the membership of -private.
If not, why not go that route again, but have this script pick up new emails from the Gapps account so google is filtering spam for us.
Because that would be horribly inefficient and a totally different approach from our current implementation?That's why I setup the Google Apps account for hp.o in the first place, hoping we could get it to work like we needed it. Unless we do a regularly-scheduled IMAP check every x minutes, there's no way we could get security@hp.o to work from Google Apps, unless we maintained a separate distribution list independent of -private... which I don't think anyone really wants.If you want to throw together a quick IMAP-checking script, I don't think anyones feelings would be hurt if we tried it out. Sean wrote this one simply because it seemed the best way to handle what we needed it to handle at the time.
Put fetchmail on a cron and use the script as the local delivery agent
for that account (via .forward, .procmailrc or whatever)?
-Matt
Correct, no more exim. If someone wants to make it work via imap or
exim, my feeling would in fact not be hurt. I was merely trying to
solve the problem, and apart from a few spams to security@ (which I'm
happy to ignore in the short term), it seems to be working.
> Running a spam filter is a usually CPU-intensive task, so if we
> could avoid sticking that on our slice that'd probably be best. If
> that's the only other option, I'd much rather use Google Apps with
> an IMAP check...
Agreed. On my box, the main resource hog is spam filtering.
S