Hardware Touch, Stronger SSH
ubicloud.comIn my opinion only, Yubico has done no favors to the Fido by their marketing. A result of trying to make Yubikey synonymous with Fido, it has become unclear what Fido does.
And as a result of how they market their keys, decisions Fido keys are presented with a cost of $20 - $60. Why $60, for a simple Fido key? Because for $60 you get not only Fido, but Flippo, Froggo, x.6s8o and more-o.
The result is that most people know the name Yubikey, but don't really know Fido, or what it is. On Amazon if you search for Fido you get mostly Yubikeys. There were other brands, but Yubico appears to have snuffed them. At one point there was an open source version that worked just as well as a name brand.
As for value? If you are a big corporate type this is the cat's meow. But otherwise? What other hardware is $60? A Raspberry Pi 4? I can get little cheap USB thingies from China at 6 for a dollar.
I am not pointing at Yubico as they have done well making profits from corporations. Rather the Fido Alliance. Looking at the Fido Alliance provides a first pass at answering the question "Who Benefits?"
https://fidoalliance.org/overview/leadership/
Perhaps it is fair to ask "What benefit" as well.
Corpocracy. You gotta love it.
Most Government organizations mandate FIPS Yubikey’s that are outrageously priced.
Yes, the $60 is clear regulatory capture. It also sets back security by raising the barrier to using these devices.
while you are right, security is generally not cheap.
you can get that $5 china fido key, but are you sure it's you who owns it?
I was recently looking for a security key, and eventually I did pay the yubico tax, because saving $20 by getting another one seemed unwise given the stakes.
>you can get that $5 china fido key, but are you sure it's you who owns it?
Seems like a moot point because it'd be very difficult for a rogue fido key to exfiltrate data. I'd be far more concerned about random chinese IOT gadgets, which most people don't have a problem with.
Hmm yes but it's possible to compromise private key generation to only create a very small predictable subset of keys. In fact some smartcards from Infineon suffered from this as a bug. And thus they can be brute forces. It requires some serious crypto chops to determine if this is the case. Obviously it's not like the first 60 bits being zero or something. And the private key is made to not be extracted in this kind of device making it even harder.
One issue i see is that it's a sealed package; it wouldn't be immediately apparent if someone added extra hardware/functionality.
More likely though I'd expect you'd just get some form of a clone device
Couldn't they ship pre-compromised? Storing the RNG seed and private key at the factory.
Devil’s advocate: How do they map that data to a user when you are buying through a maze of resellers?
they dont, they try against all the keys, there are at most a few billion of them
see Dual_EC_DRBG
It won't be as easy as that because you can generate a private key multiple times and notice it's the same.
However yes a very limited entropy in the private key is much harder to detect especially because on this kind of device you can't see the private key directly.
You're paying for brand and the fact they make key exfiltration very hard.
Getting the key out of rpi4 will be trivally easy if someone stoles it, not so much for hardware key.
I am surprised that competition didn't kept them in check, we're using them for more than a decade and the price just keeps slowly creeping in.
Run-off-the-mill smart cards have had non-extractable keys for decades. They only cost cents in manufacturing.
In raw materials, yes, but a lot of people were involved in developing and then conducting security evaluations on the MCU on the card, as well as all the software that runs on top, and those people do not work for free.
>I am not pointing at Yubico as they have done well making profits from corporations. Rather the Fido Alliance. Looking at the Fido Alliance provides a first pass at answering the question "Who Benefits?"
>https://fidoalliance.org/overview/leadership/
>Perhaps it is fair to ask "What benefit" as well.
>Corpocracy. You gotta love it.
You're really beating around the bush, trying to imply there's something shady going on, but don't articulate what it actually is. So let me ask: what's the conspiracy here? That the fido alliance is a front for an evil cabal of tech companies trying to... improve security? sell overpriced security keys?
This is not a conspiracy. This is just corporations acting in their best interest. Exactly the same as MicroSoft acted in their best interest as described here: https://en.wikipedia.org/wiki/United_States_v._Microsoft_Cor....
If that is confusing here is link to the Friedman Doctrine that explains it. https://en.wikipedia.org/wiki/Friedman_doctrine
When we see a technology that appears beneficial and is not adopted, I think it is fair to wonder why that is.
For me the key points I ponder are:
- over several years I saw articles on HN that supposedly promoted Fido, but almost always they talked about Yubikeys. This continues.
- Solokeys built an open source Fido key. They were priced very low compared to Yubikeys, but functioned just as well. You could buy them on Amazon at one point (and I did)
- the Fido Alliance Accreditation fees https://fidoalliance.org/certification/authenticator-certifi...
So no. I do not see a conspiracy, I just see an array of corporations acting according to the Friedman Doctrine.
Perhaps a good question is what benefits might those corporations gain from their actions. Would Google and Apple benefit from broad adoption of Fido keys or would it somehow lessen their profits? I don't know the answer, but I know the question.
>If that is confusing here is link to the Friedman Doctrine that explains it. https://en.wikipedia.org/wiki/Friedman_doctrine
>When we see a technology that appears beneficial and is not adopted, I think it is fair to wonder why that is.
>...
>Perhaps a good question is what benefits might those corporations gain from their actions. Would Google and Apple benefit from broad adoption of Fido keys or would it somehow lessen their profits? I don't know the answer, but I know the question.
Again, I don't see any cogent arguments here aside from a vague anti-corporations sentiment along the lines of "corporations are greedy so they must be trying to oppress us at every opportunity". You mention "I think it is fair to wonder why that is", but you haven't articulate how hobbling u2f/fido/webauthn benefits the tech giants, when security is a huge pain point for them (both for their employees and their customers), and therefore they presumably benefit from it being adopted.
>- over several years I saw articles on HN that supposedly promoted Fido, but almost always they talked about Yubikeys. This continues.
Is there any evidence this was perpetuated by the fido alliance and/or their sponsors? Should we think there's a conspiracy by github because people confuse git with github?
>- Solokeys built an open source Fido key. They were priced very low compared to Yubikeys, but functioned just as well. You could buy them on Amazon at one point (and I did)
>- the Fido Alliance Accreditation fees https://fidoalliance.org/certification/authenticator-certifi...
What is this supposed to be evidence of? If anything this disproves your point that there can be competitors to yubikey.
It's an observation, not an argument.
Using a Token2 based id_ed25519_sk_rk key, I found very helpful to configure a different `pushurl` in `.git/config`. This allows to pull via HTTPS w/o a hardware touch.
[remote "origin"]
url = https://github.com/freeCodeCamp/devdocs.git
pushurl = git@github.com:freeCodeCamp/devdocs.gitGitHub dropped http authentication so this only works for public repos (not that the UX or security of http auth for git is nice).
Can git be configured to use different keys for push and pull? (You can obviously use different upstreams, but thats not as elegant.) Most git servers let you specify read vs read-write privileges (aka “deployment keys”) so you could use one key to pull updates that doesn’t need touch and another key to push (which does).
You configure separate entries in your ssh conf.
pushurl = gh-auth:freeCodeCamp/devdocs.gitHost gh-auth Hostname github.com Identityfile blah User gitGitHub did not drop http auth. They prefer you use http instead of ssh.
What they dropped was auth using your account name and password. You need to use a token as your password or use an extra tool like their cli client to setup auth (but it sucks if you have multiple accounts).
This is how you handle it as an individual developer, but in a corporate environment things get real difficult, real fast. You need to set up your VMs and Git host to only trust certificates signed by an SSH certificate authority, and you need to work with users to submit the public key from the hardware-backed key to IT (controlling the CA) to get the public key signed and a certificate issued. Establishing trust when dealing with remote workers is hard unless you have both the budget and leadership patience to pay for overnight shipping, and even then, most people don't have access to tamper-proof packaging. Furthermore, for SSH CA support, GitHub requires Enterprise Cloud, GitLab requires Premium and self-hosted instances are not supported.
Would love to hear more from people getting this successfully set up at scale in corporate environments. I've seen big companies with lots of InfoSec talent not even attempt this.
I can't speak to actually setting it up, but where I work we have an IT-provided yubikey ssh-agent that handles getting all that stuff set up, and we just paste the public key from our individual yubikeys into our authorized ssh keys with our on-prem-hosted bitbucket server. However almost everyone I know quickly gets sick of touching the yubikey for every git remote operation and just generates their own local SSH key to use for git since doing so is not forbidden. It's definitely not High Security, but since our git is on-prem and can only be accessed from within the corporate VPN the risks are probably lower than if we were using something shared on the public internet.
The obvious solution is an ssh-agent integration that caches the touch-derived key for up to N hours or until the workstation is locked (as a proxy for user-is-away event), AND integrates with secure desktop (à la UAC) to securely show a software-only confirmation prompt/dialog for subsequent pushes within the timeout window.
(Tbh, a secure-desktop-integrated confirmation dialog would solve most issues that needed a hardware key to begin with.)
> almost everyone I know quickly gets sick of touching the yubikey for every git remote operation and just generates their own local SSH key to use for git since doing so is not forbidden
Yes, that's the exact problem at hand. If you generate your own local SSH key, the private key sits on the disk, and it can be stolen by malware (see article).
I'm asking how people set up the controls such that only hardware-based keys are signed by the CA.
If you aready have an SSH CA, why not just issue ephemeral certs lasting for several seconds or minutes? What risk would be addressed by adding hardware keys into the mix?
How do you prevent malware running on the pwned laptop from asking for an ephemeral cert to be issued? How do you know a human being is in the loop? Usually ephemeral sessions are up to 15 minutes (also to deal with misaligned clocks and unhappy users) - plenty of time for malware to ship the cert back to a command-and-control server.
This is the key advantage of hardware keys, the fact that the physical press is required prevents the keys from being exfiltrated from the machine by malware.
> How do you prevent malware running on the pwned laptop from asking for an ephemeral cert to be issued?
If you have malware capable of code execution, restricting the ability to issue one command is not going to be a meaningful control, especially with something like a physical touch which most users are just conditioned to accept, or can be trivially phished into accepting.
> plenty of time for malware to ship the cert back to a command-and-control server.
If your infrastructure cannot distinguish legitimate traffic, or you do not have a defensible network perimeter, again a physical touch is not going to be meaningful; it is not the panacea you are looking for.
I'd be fished in a heartbeat. I have to tap my key like 10 times every morning and then several times more throughout the day due to random logouts. Could be my IDE, a broken SSH connection or internal site that randomly decides to request it again and of course the popup gives no indication to where the request came from. It's ridiculous.
I think things would be more secure with fewer prompts because i wouldn't be conditioned to just tap every time it pops up.
> This is the key advantage of hardware keys, the fact that the physical press is required prevents the keys from being exfiltrated from the machine by malware.
Secure elements prevent exfiltration. Touch requirements prevent on-device reuse by local malware.
On Apple Silicon devices with macOS 26+, SSH keys can be natively stored in the Secure Enclave, protected via TouchID: https://news.ycombinator.com/item?id=46025721
It only supports sk-ecdsa-sha2-nistp256 key format, however that is widely supported currently.
Been using ed25519-sk with Yubikey for a few years now. Key is stored in KeepassXC and loaded in my SSH agent upon unlock.
It makes my SSH key pretty portable across devices
My approach aswell. Lock down ssh-agent and restrict its usage as much as possible. Securing your keys is also very reasonable but it cant silence this naging voice in the back of my head that keeps reminding me of a compromised ssh-agent or shell, whenever i authorize privileged actions.
You can also do something similar with any computer that has a TPM. It's unfortunate that people don't really know about it, but I guess the tools available aren't that user friendly
> It's unfortunate that people don't really know about it, but I guess the tools available aren't that user friendly
This is my cue.
Thank you for sharing!
I bought several "Security Key NFC by Yubico": their cheapest model, no storage or fancy stuff.
My personal strategy is to use keys generated this way:
ssh-keygen -t ed25519-sk
Rules:
- A generated key never leave the machine it was generated on.
- ssh agent is never used
- ProxyJump in HOME/.ssh/config or -J to have convenient access to all my servers.
- DynamicForward and firefox with foxyproxy extension to access various things in the remote network from my local machine (IPMI, internal services, IoT, ...)
- On the web no passkey, only simple 2FA webauthn.
My understanding is that more features including "storage" means more attack surface so by avoiding it you're 1/ more secure 2/ it's cheaper.
White paper on passkey says their security is equal to the security of the OS (Microsoft Windows ...) so I avoid passkeys.
The more expensive one works as smart card so you can both generate and keep the key as hardware only. Works for SSH and GPG too
The generated FIDO keys with "[...]-sk" are hardware-only too, the "key" you load is only an "identifier" associating the onboard passkey, allowing you to add it on multiple computers but still requiring the FIDO key present to use[1]:
> ssh-keygen(1) may be used to generate a FIDO token-backed key, after which they may be used much like any other key type supported by OpenSSH, so long as the hardware token is attached when the keys are used. FIDO tokens also generally require the user explicitly authorise operations by touching or tapping them.
> [...]
> This will yield a public and private key-pair. The private key file should be useless to an attacker who does not have access to the physical token. After generation, this key may be used like any other supported key in OpenSSH and may be listed in authorized_keys, added to ssh-agent(1), etc. The only additional stipulation is that the FIDO token that the key belongs to must be attached when the key is used.
IMO the baseline Security Key ($20) series is now enough, unless your setup uses PGP, legacy SSH that doesn't support these key types, or if you're using a real certificate for e.g. code signing.
1: https://www.openssh.org/txt/release-8.2#:~:text=The%20privat...
This is how I've been doing it: https://stosb.com/blog/using-openpgp-keys-for-ssh-authentica...
Slightly different as I generate a PGP key on the computer and then load it to the Yubikey, which means I can have backup keys with the same secret keys.
I never really got "touch to use" working though, if anyone knows how to do it with GPG keys I'd really appreciate it!
Shameless link to my own blog where I use a yubikey to store my SSH private key, a long with some advice for use in macos and Linux.
https://stephentanner.com/ssh-yubikey.html
Hopefully someone finds it useful.
The biggest issue I ran into was when folks wrote some tools that rely on ssh sock auth to automate connection to remote boxes. Not fun if you have to tap for every box.
Filler pr jippo fluffer article aside, anyone tried to self host ubicloud lately? A year and a half ago it was super cumbersome, wondering if I should give it a new try now.
Could someone please explain to me why this would be better than storing the key with 1Password and biometric authentication?
SSH using GPG Yubikeys and git signing using GPG was quite a process to set up on Windows a few years ago. Not something I'd want or know how to repeat. Hopefully things have improved in the mean time.
You can use SSH keys for signing now, so you don't need GPG at all.
ssh-add -c (confirm) can somehow mitigate the "misuse of ssh-agent in the background" the article is talking about
How will this work with agents?
You just get a prompt.
Problem is really there is no good way for the prompt to have the name of actual app that asked when it is forwarded.
That's the neat part, it doesn't
Now that's just incorrect