Charger8232

joined 1 year ago
 

This post is long and kind of a rant. I don't expect many to read the whole thing, but there's a conclusion at the bottom.

On the surface, recommended security practices are simple:

  • Store all your credentials in a password manager
  • Use two factor authentication on all accounts

However, it raises a few questions.

  • Should you access your 2FA codes on the same device as the password manager?
  • Should you store them in the password manager itself?

This is the beginning of where a threat model is needed. If your threat model does not include protections against unwanted access to your device, it is safe for you to store access your 2FA codes on the same device as your password manager, or even in the password manager itself.

So, to keep it simple, say you store your 2FA in your password manager. There's a few more questions:

  • Where do you store the master password for the password manager?
  • Where do you store 2FA recovery codes?

The master password for the password manager could be written down on a piece of paper and stored in a safe, but that would be inconvenient when you want to access your passwords. So, a better solution is to just remember your password. Passphrases are easier to remember than passwords, so we'll use one of those.

Your 2FA recovery codes are something that are needed if you lose access to your real 2FA codes. Most websites just say "Store this in a secure place". This isn't something you want to store in the same place as those (in this case our password manager), and it's not something you will access often, so it's safe to write it down on a piece of paper and lock it in a safe.

Good so far, you have a fairly simple system to keep your accounts safe from some threats. But, new problems arise:

  • What happens if you forget your master passphrase?
  • What happens if others need access to your password manager?

The problem with remembering your passphrase is that it's possible to forget it, no matter how many times you repeat it to yourself. Besides naturally forgetting it, things like injuries can arise which can cause you to forget the passphrase. Easy enough to fix, though. We can just keep a copy of the passphrase in the safe, just in case we forget it.

If someone else needs to access certain credentials in your password manager, for example a wife that needs to verify bank information using your account, storing a copy of the password is a good idea here too. Since she is a trusted party, you can give her access to the safe in case of emergencies.

The system we have is good. If the safe is stolen or destroyed, you still have the master passphrase memorized to change the master passphrase and regenerate the 2FA security codes. The thief who stole the safe doesn't have your password manager's data, so the master passphrase is useless. However, our troubles aren't over yet:

  • How do you store device credentials?
  • How do you keep the password manager backed up?

Your password manager has to have some device in order to access it. Whether it's a phone, computer, tablet, laptop, or website, there needs to be some device used to access it. That device needs to be as secure as your password manager, otherwise accessing the password manager becomes a risk. This means using full disk encryption for the device, and a strong login passphrase. However, that means we have 2 more passwords to take care of that can't be stored in the password manager. We access those often, so we can't write them down and store them in the safe, Remembering two more passphrases complicates things and makes forgetting much more likely. Where do we store those passphrases?

One solution is removing the passwords altogether. Using a hardware security key, you can authenticate your disk encryption and user login using it. If you keep a spare copy of the security key stored in the safe, you make sure you aren't locked out if you lose access to your main security key.

Now to keep the password manager backed up. Using the 3-2-1 Backup Strategy. It states that there should be at least 3 copies of the data, stored on 2 different types of storage media, and one copy should be kept offsite, in a remote location (this can include cloud storage). 2 or more different media should be used to eliminate data loss due to similar reasons (for example, optical discs may tolerate being underwater while LTO tapes may not, and SSDs cannot fail due to head crashes or damaged spindle motors since they do not have any moving parts, unlike hard drives). An offsite copy protects against fire, theft of physical media (such as tapes or discs) and natural disasters like floods and earthquakes. Physically protected hard drives are an alternative to an offsite copy, but they have limitations like only being able to resist fire for a limited period of time, so an offsite copy still remains as the ideal choice.

So, our first copy will be on our secure device. It's the copy we access the most. The next copy could be an encrypted hard drive. The encryption passphrase could be stored in our safe. The last copy could be a cloud storage service. Easy, right? Well, more problems arise:

  • Where do you store the credentials for the cloud storage service?
  • Where do you store the LUKS backup file and password?

Storing the credentials for the cloud storage service isn't as simple as putting it in the safe. If we did that, then anyone with the safe could login to the cloud storage service and decrypt the password manager backup using the passphrase also stored in the safe. If we protected the cloud storage service with our security key, a copy of that is still in the safe. Maybe we protect it with a 2FA code, and instead of storing the 2FA codes in the password manager, we store it on another device. That solves the problem for now, but there are still problems, such as storing the credentials for that new device.

When using a security key to unlock a LUKS partition, you are given a backup file to store as a backup for emergencies. Plus, LUKS encrypted partitions still require you to setup a passphrase, so storing that still becomes an issue.

Conclusion

I'm going to stop here, because this post is getting long. I could keep going fixing problems and causing new ones, but the point is this: Security is a mess! I didn't even cover alternative ways to authenticate the password manager such as a key file, biometrics, etc. Trying to find "perfect" security is almost impossible, and that's why a threat model is important. If you set hard limits such as "No storing passwords digitally" or "No remembering any passwords" then you can build a security system that fits that threat model, but there's currently no security system that fits all threat model.

However, that doesn't let companies that just say "Store this in a secure place" off the hook either. It's a hand wavy response to security that just says "We don't know how to secure this part of our system, so it's your problem now". We need to have comprehensive security practices that aren't just "Use a password manager and 2FA", because that causes people to just store their master passphrase on a sticky note or a text file on the desktop.

The state of security is an absolute mess, and I'm sick of it. It seems that, right now, security, privacy, convenience, and safety (e.g. backups, other things that remove single points of failure) are all at odds with each other. This post mainly focused on how security, convenience, and safety are at odds, but I could write a whole post about how security and privacy are at odds.

Anyways, I've just outlined one possible security system you can have. If you have one that you think works well, I'd like to hear about it. I use a different security system than what I outline here, and I see problems with it.

Thanks for reading!

 

Yesterday I decided to start "officially" selfhosting. With almost no experience with Docker, I struggled for eight hours straight, but I finally have it working.

Currently, the two tools I am selfhosting with Docker Compose are LibreTranslate and spotDL. I'm only accessing them over the local network using a direct IP:PORT, so there's no domain name. I don't want to use a custom DNS, since it is fingerprintable online, so I want to keep it the same as my VPN.

With that said, I want to add encryption to the connections. I was able to generate my own self signed certificates with this command:

sudo openssl req -x509 -nodes -days 365 -newkey rsa:2048 -keyout ./certs/key.key -out ./certs/cert.crt

spotDL was easy to setup with these self signed certs, since it has command flags for --enable-tls, --key-file, and --cert-file. LibreTranslate has an environment variable for - LT_SSL=true, however it gives the following error:

libretranslate  | (URLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1006)')),)

That led me to this issue which is 2 and a half years old. LibreTranslate doesn't have a way to specify certificates that I know of.

I tried using Nginx Proxy Manager to create a reverse proxy, but I couldn't quite figure it out. (I also didn't understand Docker Compose at the time. I had a few hours to go before I did) NPM also seems to want Let's Encrypt certificates which can't be given without a DNS record. I tried manually providing certificates in the config file, but I don't quite understand NPM enough to be able to set it up properly on my own.

My requirements are:

  • No changing the DNS from my VPN's default
  • No port forwarding, everything should be accessed by the local network only
  • No email required (ability to use a fake email without risk is fine)
  • Only free and open source software
  • Modern security standards where available

I also would like help adding the self signed certificates as a permanent exception in Brave browser, if possible.

 

This question has been answered. Please stop trying to repeat information that has already been said many times before. Everything in this thread is in good faith, I am here to learn, so I will make mistakes. Furthermore, if you want to contribute something new, please read the entire post to avoid misunderstanding the purpose of this post.

Selfhosting is useful when you either need a lot of storage or a lot of processing power. For example, Kiwix is useful to selfhost on a server because a lot of its content can take up terabytes of storage, which a phone may not have. LLMs are also useful to selfhost because they require a degree of processing power that, again, a phone may not have.

In both cases, there is also a need for perpetual access. If you simply hosted an LLM on your home computer, it wouldn't be very useful to access from your phone since your computer won't be running all the time. So, a separate always-on server is needed.

However, there are some selfhosted software that I don't see a use for. For example, Immich. Immich requires to be run on a server to function, but a lot of (or even all) of its functions are things that could reasonably done entirely on-device. Aves combined with some automatic backup solution such as Nextcloud gets (from what I can tell) most of the functionality Immich offers. Obviously, some features like AI image tagging are missing, but you get the point. AI image tagging is also something that could be run on-device as well, since it's mostly lightweight (iPhones are capable of it). Having a setup like that also comes with the benefit of automatic backups being completely optional, rather than required.

There's no reasonable need for extra storage or extra processing power needed for that use case, from what I can tell. (Disclaimer: I haven't actually used Immich before, so this is speculation. I apologize if I'm missing something obvious) There's a lot of other selfhosted tools like spotDL which have a selfhosted web UI, but no GUI that can be installed outside of a web browser.

I guess my question is why there are so many selfhosted tools that unnecessarily require being run on a separate device. I do understand the legitimate use cases some of them have, but others seem better off on-device airgapped. This especially became an issue trying to find a notes app for Android that requires no account and runs fully locally, or an RSS reader that loads from the device itself. I found Joplin and Feeder or Read You as the software for each of those. I don't like "server-based" selfhosting for things that could be done from the device itself.

I'm sorry if this turned into a rant. If someone could help me understand, I would appreciate that very much.

Cheers!

Edit: The comparison here isn't between selfhosting and using a cloud provider. The comparison here is between selfhosting on a server and running explicitly on-device (besides where extra storage or processing power is required)

Answer

So that nobody has to dig through the comments for answers, this is what I've learned: In the case of Immich, its purpose isn't designed to be a photo gallery. It's designed to be a more polished backup solution, designed explicitly for photos and not general files. While Nextcloud could be used to backup photos, it's not as focused on photos as Immich, and so it isn't as nice to use for that purpose. Immich also allows you to share photos with a link, rather than relying on a cloud provider to do that for you. There's also another benefit to selfhosting that I hadn't entirely realized, which is availability across devices. Some things like an eBook library may not take up much space, but it's convenient to not have to sync manually (or automatically) across devices, and instead access it from a central server. That same logic is true for RSS readers as well, since it's inconvenient to manually add and sync feeds across devices. Syncing across devices can be done with something like Syncthing in some cases, but not all, and so that's where selfhosting can be useful.