Put your SSH keys in your TPM chip
raymii.org
raymii.org
I don’t see much difference between that and storing the key on a TPM. If you have one key and you lose access to that key, then you lose access to the server.
Point: you need a backup key anyway.
Just paste all of your devices' public keys into your authorized_keys file and leave a comment at the end for what device it's for. in Userify, it literally goes right into your nodes' authorized_keys file almost verbatim. (disclaimer: I work at https://Userify.com)
And then, if you leave your token or laptop at the airport or whatever, just remove that key right from your phone and it'll take effect in seconds across all the nodes/instances (if you're using Userify) or you can just write a quick for-inline-sed loop to remove it from your authorized keys everywhere.
OTOH for SSH use if you lose the key you just create a new one, it's not like you've lost the only copy of your Bitlocker key.
I don't think average Joe is going to understand these passkeys either.
With a password, you open your password manager, copy the password in memory, paste it into the input field and trust that nobody could read it from your clipboard and that the program handling the password does it correctly. If your password leaks on the way, it's leaked.
With FIDO2, the server sends a challenge and asks your HSM (or TPM, not sure what the right word is) to sign it with your private key. So the server can verify that you own the private key, but if the challenge or the response leaks, it's just this one time. Next time it will be a new challenge.
Also for the average Joe, the result is that the "passkey" is the fingerprint or the face recognition and there is no password. It feels like they have only one password: the biometry/face recognition (or a master password, I guess?). So passkeys are superior to passwords in that sense.
Fun fact 1: some people hate passkeys because they don't want to be forced to rely on TooBigTech for them. Currently I use my Yubikeys as passkeys everywhere and it works well, so I do NOT depend on TooBigTech.
Fun fact 2: FIDO2 on current Yubikeys (and HSM in general, I think) tend to use classic cryptography which would be broken by quantum computers. A password used with symmetric encryption is not broken by quantum computers. So there may be a period of time where this becomes a tradeoff (you may have to decide whether the most likely attack is a quantum computer breaking your authentication or a malware stealing your password)?
I don't do that. My password manager simulates keystrokes 2 seconds after I hit the button. I switch to the other window and my password gets punched in without going through the clipboard. Specifically to avoid this attack.
> Currently I use my Yubikeys as passkeys
I have Yubikeys but for 2FA. So we're back to 1FA now but just "something you have" and no "something you know" ?
So a malware on your computer can just listen to the keystrokes, or read on the screen? If the OS is compromised, they can extract the password. With a passkey they can't.
> So we're back to 1FA now but just "something you have" and no "something you know" ?
You can set up a PIN on your Yubikey, so that's "something you have" and "something you know", and you can request physical presence ("touching the yubikey") on top.
Not quite. First of all, passkey is not a standardized term. But usually it refers to a key that can be used for authentication on its own, not as a 2FA along with a password.
A FIDO2 key can be a passkey, or not, depending on the service or configuration.
FIDO2 and WebAuthn added some fields necessary to make this work "securely", e.g. asking the key to verify the user separately (e.g. a PIN, which serves as a second factor), or asking the key whether it is device bound or roaming, so individual sites/enterprises can enforce their security policies
However not all devices play well with it, e.g. iOS and Android don't ask 1pass for the passkey. I also couldn't make it ask NFC for my hardware Yubikey with passkeys, but maybe I just did something wrong.
It's false that passkeys cover biometrics. They cover password + OTP (aka 2FA aka MFA, although BestBuy requires OTP even after logging in with a passkey).
As someone who stores my SSH keys in my TPM, and has struggled with picking the right PCR values for Secure Boot in the past, I'm interested in learning more.
This may be bash-only, but a space before the command excludes something from history too.
Personally I like this which reduces noise in history from duplicate lines too. export HISTCONTROL=ignoreboth:erasedups
I can't find it now, but I believe someone from Tailscale commented on HN (or was it github?) on what they ran into and why the default was reverted so that things were not stored in the TPM.
EDIT: just saw the mention in the article about the BIOS updates.
https://github.com/tailscale/tailscale/issues/17622
https://news.ycombinator.com/item?id=46532666 (direct comment link, more discussion on the issue in the parent)
In theory the Linux kernel keyring would help here, even with a tsm or in conjunction with it.
Unfortunately as the industry abandoned the core Unix permission system (uid/gid) all of these methods just get a devfs[null] bind mount.
Only process that also support the traditional co-hosting model like nginx and Postgres do.
We would need nonce keys to gain no value from kernel memory or hardware storage.
You just need to tighten your sshd config, you can even add a "touch required" of the Yubikey to the sshd config. Has been in debian stable since like 11 I think?
So it's super friendly to integrate and very secure, as you need to physically be on your pc, have your yubikey and have your exact pc. So that's a lot of factors.
Token-based keys, to tptacek's point, is that they can be a giant pain once you start scripting across fleets.
And keys cannot be stolen from backups.
Or stolen without your knowledge when you left your laptop unguarded for 5min.
Not every attacker has persistent undetected access. If the key can be copied then there's no opportunity for the original machine's tripwires to be triggered by its use. Every second malware runs is a risk of it being detected. Not so, or not in the same way, with a copied key.
This is really cool and goes beyond the usual steps of securing the key, but handling "what you see is what you sign" and key usage user confirmation at the OS level, which can be compromised much more easily (both input and output).
Quote: "Android Protected Confirmation is deprecated due to the high support/maintenance cost for Android device makers and low adoption rate among app developers. APC requires Android device makers to have a substantial amount of device-specific UI code running in the trusted execution environment. That has proven to be expensive to maintain and non-scalable, as there cannot be a single implementations device makers can share or use as a reference. Additionally, app developers have not adopted this feature, as the Android platform offers other mechanisms for authentication a user's intent. These mechanisms, such as authentication-bound Keystore keys, are less secure than Trusted UI, but are more wide-spread. While we explore alternatives to APC that are viable to the device makers ecosystem, we sunset the APC API."
Still, I think this was one of the most ambitious and user-beneficial implementations of trusted computing I've seen so far, in that it theoretically safely allows a completely rooted/user-owned device to still participate in things like online banking or e-government transaction authorization. I hope it'll return in some form.
A laptop and a phone are both general purpose computers with "TPM chips", so "you could implement that on android" is as true as "you could implement that on a white computer".
There was something about Macs. It took them a while to get a TPM. But I think now they do, so macs can do it too.
There are many ways to implement this. I think some Chromebooks have FIDO gated on a physical button.
If you have an unlocked device with keys usable requiring a mere touch, I'm not sure fingerprint adds much value. A button would be enough.
Actually checking with fingerprint only addresses an extremely narrow attack where someone who wants to attack you steals your device (so already physical access, meaning not DPRK hackers) while it's unlocked, and only getting a window of opportunity until you've called your security department to lock your account. … and yet this attacker would NOT be willing to use force against your person, to make you use your fingerprint.
Sure, if that's a threat model that's worth your time, use fingerprint too.
Keep in mind that already going from software only (and arguably this includes OTP app on your phone) already means effectively going to zero. Google moved to security keys and says “We have had no reported or confirmed account takeovers since implementing security keys at Google” — https://krebsonsecurity.com/2018/07/google-security-keys-neu...
So there are extreme diminishing returns after just security key with touch.
An app solution even gets a callout in that article as being not as good.
Of course a real secure attention sequence would be preferable, such as e.g. requiring a Touch ID press on macOS for keys stored in the Secure Enclave. Not sure if TPM supports something similar when a fingerprint sensor is present?
The PIN can be an arbitrary string (password).
In addition to biometric authentication, Windows Hello supports authentication with a PIN. By default, Windows requires a PIN to consist of four digits, but can be configured to permit more complex PINs. However, a PIN is not a simpler password. While passwords are transmitted to domain controllers, PINs are not. They are tied to one device, and if compromised, only one device is affected. Backed by a Trusted Platform Module (TPM) chip, Windows uses PINs to create strong asymmetric key pairs. As such, the authentication token transmitted to the server is harder to crack. In addition, whereas weak passwords may be broken via rainbow tables, TPM causes the much-simpler Windows PINs to be resilient to brute-force attacks.[139]
So you see, Microsoft needs a way to describe an access code that isn't a password, because it's more secure than that, but yet it isn't exactly a number, so what do you call it? "PIN" is perhaps an unfair recycling of an in-use term, but should they coin a neologism instead? Would that be less confusing?
Communicates it is meant to be secret, and can be a short memorable thing.
If your computer is compromised after you've already entered the PIN, or there is an app running on the computer but it is not sufficiently privileged to sit in between you and the TPM, no.
That's quite good protection generally. The defense against this type of attack is to get a smartcard reader with an on-board PIN entry keypad - those do exist, but it's quite a step.
Well no thanks, that risk is much higher than what this is worth.
There's also the TPM speed issue. My computer takes ~500ms to sign with an ECC256 key with the TPM, which starts to become an issue when running scripts that use git operations in serial. This is a recurring problem that people tend to blame on export controls: https://stiankri.substack.com/p/tpm-performance
I put my ssh keys into the Mac’s TPM and now it asks for a password/touch ID when I use it.
Unfortunately I forget what commands I used
If you are doing something illegal or controversial with the key, then yes it would be foolish to store it in the cloud.
If your main concern is it becoming compromised due to a local exploit or physical breach, then I'd argue it is a strong option.
"Not trusting a private company" does not equate to "doing something illegal or controversial", though.
Since it just uses PKCS#11, it also works with tpm_pkcs11. Source for the various bits that are bundled is here [1].
Here's an overview of how it works:
1. Application asks to sign with GPG Key "1ABD0F4F95D89E15C2F5364D2B523B4FDC488AC7"
2. GPG looks at its key database and sees GPG Key "1ABD...8AC7" is a smartcard, reaches out to Smartcard Daemon (SCD), launching if needed -- this launches gnupg-pkcs11-scd per configuration
3. gnupg-pkcs11-scd loads the SSH Agent PKCS#11 module into its shared memory and initializes it and asks it to List Objects
4. The SSH Agent PKCS#11 module connects to the SSH Agent socket provided by Keeta Agent and asks it to List Keys
5. Key list is converted from SSH Agent protocol to PKCS#11 response by SSH Agent PKCS#11 module
6. Key list is converted from PKCS#11 response to gnupg-scd response by gnugpg-pkcs11-scd
7. GPG Reads the response and if the key is found, asks the SCD (gnugpg-pkcs11-scd) to Sign a hash of the Material
8. gnupg-pkgcs11-scd asks the PKCS#11 module to sign using the specified object by its Object ID
9. PKCS#11 module sends a message to Secretive over the SSH Agent socket to sign the material using a specific key (identified by its Key ID) using the requested signing algorithm and raw signing (i.e., no hashing)
10. Response makes it back through all those same layers unmodified except for wrapping
(illustrated at [2])
[0] https://github.com/KeetaNetwork/agent
[1] https://github.com/KeetaNetwork/agent/tree/main/Agent/gnupg/...
I'm wondering why that doesn't apply here. The TPM holds the key to the cipher that is protecting your private keys. Someone uses some kind of RCE or LPE to get privileged access to your system. Now it sits and waits for you to do something that requires access to your SSH keys. When you do that you are expecting whatever user prompts come up, the malware rides along this expectation and gets ahold of your private SSH keys and stores them or sends them off somewhere. I'm not even positive that they need high degree of privileges on your box, if they can manipulate your invocation of the ssh client, by modifying your PATH or adding an ssh wrapper to something already in your path, then this pattern will also work.
What am I gaining from using this method that I don't get from using a password on my ssh private key?
Further promises are RTC that can prevent bruteforce (forced wait after wrong password entry) or locking itself after too many wrong attempts.
A good MCU receives the challenge and only replies with the signature, if the password was correct. You can argue that a phone with a Titan security chip is a type of TPM too. In the end it doesn't matter. I chose the solution that works best for me, where I can either only have all keys in my smart card or an offline paper wallet too in a fireproof safe. The choice is the user's.
Technically, a private key that was imported (and is marked as exportable) to a PKCS#11 device can subsequently be re-exported (but even then, during normal operation the device itself handles the crypto), but a key generated on-device and marked as non-exportable guarantees the private key never leaves the physical device.
https://wiki.archlinux.org/title/SSH_keys#Storing_SSH_keys_o...
And even the password can be forced to be re-entered by the agent for every use, if that level of security is wanted.
Depending on which authenticator app (or maybe applies to all?), that data either is, or can be, backed up.
A yubikey cannot be cloned.[1]
> the malware rides along this expectation and gets ahold of your private SSH keys and stores them or sends them off somewhere.
Ah, this is where your misunderstanding lies. No, the crypto operation runs ON the TPM or yubikey. The actual secret key NEVER lives in RAM. (ehem, after it was imported, if importing is the method by which it was generated)
[1] You know what I mean. Of course in principle it can be. But not like a phone where it can literally be sent via scp.
The recommended usage of a yubikey for ssh does something similar as otherwise your key consumes one of the limited number of slots on the key.
Yes, with TPM and yubikey you have the option to store the per key material on disk, encrypted by the TPM. But the way this is then used is that the PKCS software sends that encrypted blob AND the requested operation, and gets only the output back. The CPU doesn't get the SSH private key back. Just the output of the RSA operation using the key.
What am I missing?
Thank you for your reply.
(Others in the thread have confirmed that both resident and non-resident keys never leave the hardware. If you generate one that requires touch, they're fairly secure - you need physical presence and confirmation for every operation.)
It does mean that they can't use the key a thousand times. But once? Yeah sure.
That doesn’t do anything at all.
1. If the attacker is redirecting you to a different host then ssh will simply refuse to connect due to known_hosts (I guess they could have added to that file too, redirect you to a honeypot and then hopefully you’ll run “sudo” before realizing but then at that point just hijack “sudo” itself in the local machine)
2. If the attacker is trying to let you connect and eavesdrop your connection to still credentials then that also still doesn’t work as the handshake for ssh is not vulnerable to replay attacks
The attacker could trick you into signing something I guess but then that still doesn’t do anything because secrets are not divulged at any point
I guess if the yubikey is also used for `sudo` then your attack makes more sense, as the attacker could prompt you to authenticate a sudo request when you call the evil `ssh`
1. attacker wants to use your yubikey-backed ssh key, let's say for running ssh-copy-id once with their own key so they can gain access to your server
2. thus they need to trick you into touching the key when they run that command
3. the best way to trick you is to wait until you do something where you'd normally need to touch that key yourself
4. so they alias ssh to a script that detects when you're trying to connect to this server yourself, and invoke ssh-copy-id instead, which prompts you to touch the yubikey and you do
5. spit out a reasonable looking error (something that makes you think "bloody DNS, it's always DNS, innit" or something silly like that); then they undo the alias so you succeed on the next try and suspect nothing
But they won't get your private key.
Indeed, that was my point exactly a couple posts up the thread. :-)
> you may realise that something wrong happened
I think I can iterate on the exact mechanics to make this less likely. I mean it's getting off-topic but the one thing that comes to mind is to enable ControlMaster for all ssh connections which allows any second ssh invocation to skip the auth and just re-use the existing connection. ssh-copy-id is near instant then and doesn't ask anything.
At that point you might—rightly so—argue that they're no longer tricking the user into authorising a different operation. Just a reminder that if someone can run code as your local user, they can easily and sneakily gain access elsewhere. Even if you need a yubikey touch to connect there.
The original attack idea of timing the yubikey touch for when you normally expect to touch it might still be relevant for a scenario like ssh-agent forwarding to a malicious box. They can't run code as your local user, but can still perhaps trigger the agent to interact with the yubikey. Maybe.
And the best thing is that you can create several different ssh keys this way, each with a different password, if that's something you prefer. Then you need to type the password _and_ touch the yubikey.
And they won't be replacing these just because they're missing FIDO. And they can't "just" be upgraded because they aren't necessarily just Linux boxes in a trenchcoat. Nor are they necessarily running any version of OpenSSH.
These work flawlessly with the KeepassXC ssh-agent integration. My private keys are password protected, saved securely inside my password vault, and with my ssh config setup, I just type in the hostname and tap my Yubikey.
https://www.stavros.io/posts/u2f-fido2-with-ssh/
We've got private Git repos only accessible through ssh (and the users' shell is set to git-shell) and it's SSH only through Yubikey. The challenge to auth happens inside the Yubikey and the secret never leaves the Yubikey.
This doesn't solve all the worlds' problem (like hunger and war) but at least people are definitely NOT committing to the repo without physically having access to the Yubikey and pushing on it (now ofc a dev's computer may be compromised and he may confirm auth on his Yubikey and push things he didn't meant to but that's a far cry from "we stole your private SSH key after you entered your passphrase a friday evening and are now pushing stuff in your name to 100 repos of yours during the week-end").
"Your SSH keys" aren't really part of that threat model. "You" know the device you're connecting from (or to, though generally it's the client that's the mobile/untrusted thing). It's... yours. Or under your control.
All the stuff in the article about how the TPM contents can't be extracted is true, but missing the point. Yes, you need your own (outer) credentials to extract access to the (inner) credentials, which is no more or less true than just using your own credentials in the first place via something boring like a passphrase. It's an extra layer of indirection without value if all the hardware is yours.
TPMs and secure enclaves only matter when there's a third party watching[1] who needs to know the transaction is legitimate.
[1] An employer, a bank, a cloud service provider, a mobile platform vendor, etc... This stuff has value! But not to you.
What on earth do you think I make my users present keys for???
You know all those guides saying "you should never copy an ssh private key over the network. Make a new one for each device" that every idiot dev ignored? Now I can enforce that.
Not a chance. It is my key.
So does a pass phrase though, with significant less complexity and fragility.
Again, the linked article and responses here are making IMHO a pretty bad mistake with threat model analysis.
Can you explain why securing the ssh keys on a host that was fully compromised like that is anything but theater? Fine, you can't get the key out. You can just run the command directly.
Again, there are use cases where TPMs provide value to authenticate specific devices. But they are not and never have been about "keeping secrets". Your secrets are trash once the device is compromised.
I absolutely agree that they do zilch to protect your SSH keys. Hardware security keys that need physical confirmation of presence are much better for that use-case.
Which is what SSH keys are for?
The advantage of this approach is that malware can't just send off your private key file to its servers.
The use case is ssh keys! If malware can run an ssh command on the remote host, it doesn't need to steal your key, it can just install itself there. Or add its own keys to the access, etc... At best, you'd have to detect and fix that sort of thing with auditing and control, something that's isomorphic to the "third party" requirements I was mentioning.
To repeat the third time: this is all terrible threat model analysis. TPMs do not have value for individuals managing access between trusted devices. TPMs are for third-party validation.
But you just admitted that it prevents the key from being stolen, right? So the value is that the key cannot be stolen. Doesn't mean that a malware cannot use it of course. Just that it cannot extract it. Which is better than a malware extracting it.
Keep a CA (constrained to your one identity) with a longish (90 day?) TTL on the TPM. Use it to sign a short lived (16h?) keys from your TPM, use that as your working key.
But, if you are making a lot of x509 authenticated calls directly, then the speed and not needing to touch the key are important. Or if you need to ssh to 10,000 hosts quickly, things like that.
that's not good
I saw a write up where someone successfully got sshd to use a host key from a fido2 yubikey without touch, but I can't find it...
As far as "TPM vs HSM", it is soooo much simpler to make a key pair with a fido2 hardware key:
ssh-keygen -t ed25519-sk -O resident -O verify-required -C "your_email@example.com"