4 ms·
> That’s right, we need to enter the key when the server boots. Wait, but what happens with a power failure? Are all requests with page views to your server fai
by teddyh 8y ago
> That’s right, we need to enter the key when the server boots. Wait, but what happens with a power failure? Are all requests with page views to your server failing after a reboot?
> Then we found Dropbear, a very small SSH program, that you can run via the initial ramdisk (initramfs). This means we are able to allow external connections via SSH. We don’t have to fly to Iceland to boot our server, yeah!
Or you can use Mandos to be able to sleep through a reboot: https://www.recompile.se/mandos https://www.recompile.se/mandos
Disclosure: I am the co-author of Mandos.
- kitotik 8y agoOn first glance this looks great and solves a real problem. Not sure how I’ve missed this. Is there much of a plug-in ecosystem for this to support various auth backends? *edit grammar
- teddyh 8y agoI’m not sure what you mean by “auth backend”. By default, the client merely has to prove that it possesses the correct key to be given the secret encrypted password from the Mandos server. This can be changed by configuring the server not to automatically “approve” clients, which makes the whole procedure controllable over D-Bus, opening the way for any process you like to do whatever it needs to do in order to authenticate a client before the secret encrypted password is sent.
- kbenson 8y agoWas fairly leery of how useful I thought this would be, but you reference the FAQ[1] early in what I assume is an attempt to head this off, and it delivers adequately. Despite being predisposed going in thinking it was probably not worth the janky stuff you would have to do to get it working, the level of detail and thorough covering of some fairly complex attack vectors leaves me thinking that while not as secure as no automatic boot method for encrypted disks (obviously), it would definitely be a step up over not encrypting and something I would recommend without worrying that I was increasing the burden of managing the systems to a level an organization would be unprepared to deal with. Not that I'm some professional security researcher or anything, but congratulations, color me impressed. 1: https://www.recompile.se/mandos/man/intro.8mandos https://www.recompile.se/mandos/man/intro.8mandos
- cf498 8y agoI am not sure that the offset in security is that small. The assumption, that your servers will only be evaluated month later doesnt hold true in most cases. Attacking running machines quickly does happen. If your threat scenarios are little green men, they are unlikely to just shut down your server and haul it to some dark room to interrogate later. They will of course try to get the key from a running system, but thats also something you should most definitely have in your threat assessment. edit: blue men*. Forgot that green police uniforms are an outdated local thing.
- jimktrains2 8y agoRe your edit, I always thought lgm was a reference to military garb, not police. Also, many police wear black now.
- cf498 8y agoMakes sense, things got really butchered in the translation. Our police used to be green until, what feels like really recently. So there are still a lot of sayings around involving "the people in green". Now they are apparently blue across the EU.
- coldacid 8y agoTo me, "little green men" meant aliens.
- mark212 8y ago“Little green men” to me is a reference to the Russian army units that invaded Ukraine without insignia for deniability.
- cf498 8y agoThats a rather recent one however. The saying is rather old.
- jimktrains2 8y ago
- ynx 8y agoWow, that looks really cool. However, I use CentOS 7 and that documentation appears to indicate it's not an officially supported distro. Are you aware of alternatives to Mandos (even if they aren't as good) which have official support of RHEL/CentOS?
- teddyh 8y agoSorry, no idea; haven’t heard of any. Mandos is made for Debian for the simple reason that it’s what we use ourselves, and at the time Mandos was created, only Debian had something even approaching the flexibility of initramfs-tools. Nowadays with dracut, it should be possible to port it, but we haven’t looked at it.
- sneak 8y agoWhy does the root need to be encrypted? What's wrong with an unencrypted root and /var and /home and swap and /mnt/storage being on encrypted partitions/zpools? I boot my machines and then ssh into them with a script that pipes a local gpg -d to them to unlock their disks, mount appropriate filesystems, and start necessary daemons. Unless the daemons are poorly-behaved, no data that is specific to me aside from the host's SSH key is ever written to the root - only the OS/distro files. I thought this is what everyone does.
- mynegation 8y agoThat will work if you are sure that all the sensitive information you want to protect goes exclusively to the encrypted file system. Which is a tall order and security headache by itself.
- teddyh 8y agoThe root file system still, at the very least, records access times of all files, so an analysis of the the root file system releals, for all programs, when you last ran them, and for all libraries, when they were most recently loaded, etc. With Linux defaults, IIRC this timestamp has a limitied granularity of a day, but this can be telling enough.