framefilter/keyroost v0.15.0 on GitHub

Added

Import a key and certificate from a .p12/.pfx file. keyroostctl piv key import --slot 9a --in FILE writes a private key and its certificate
from a PKCS#12 file into a PIV slot, then tests the slot as piv test
does; in the app, it's "Import key from file…" on a PIV slot.
Legacy-encrypted files (RC2/3DES, the kind some certificate authorities
such as Spain's FNMT still issue) open directly. keyroost checks that the
key can take the file before it asks for the management key. A file with
several keys is listed for --entry N, and --chain-out FILE saves the
authority certificates, which are not written to the card. On a
Nitrokey 3, RSA keys go into slot 9A; HID Crescendo cards aren't supported
yet. An imported key can't be attested as generated on the key. (#174)
fido info --reader reads a key through a smart-card reader, as
fido reset --reader already could. A key that only speaks U2F there gets
the same line as over USB, in the app too, instead of an error. On Windows,
where only a program running as administrator may reach a key's FIDO2 side
through its reader, a refusal says so. With no options, fido info now
picks the key the same way fido reset does, so with several FIDO keys
(including one only reachable through a reader) it asks which one. (#215)
doctor and the app say when a key's smart-card part has no reader
(Linux). When a key has a smart-card interface but the smart-card service
lists no reader for it, usually because the system's smart-card driver
doesn't know the model yet, doctor says so and the app adds a line to the
key's Overview page. (#215)
openpgp info notices a key whose fingerprint is missing. It now
reads each slot's public key (no PIN) and, when a slot holds a key but no
fingerprint, says (none), but a key is present; the app's OpenPGP card
says the same, (none), but a key is present. --json gains
key_present_sig, key_present_dec and key_present_aut (true, or
null when the card doesn't tell). (#215)
piv --json info says whether the PIN is verified and whether each slot
holds a key. New pin_verified (true/false) and, per slot, key_present
(true, false, or null when the key can't tell). piv info and the app's
PIV header now say "PIN verified" instead of "PIN accepted", matching
openpgp pin verify. (#215)
piv info lists the retired key slots that are in use. Slots 82–95
show after the standard four when they hold a certificate or a key, in
the text output and in --json's slots. (#221)

Changed

No PIN or touch policy for new PIV keys on a Nitrokey 3. On firmware
newer than 1.8.3, up to 1.9.1, piv key generate and the app's Generate
dialog used to warn about a PIN or touch policy and then send it.
According to Nitrokey's published PIV source, this firmware doesn't apply
those settings and has nowhere to store one; its PIN rule is fixed per
slot and it has no touch requirement. keyroost now
refuses a policy other than default on these keys, as it already did on
older firmware; --force in the CLI or Enable Anyway in the app still
sends it. (#174)
Authenticator codes come back in one read. The Authenticator tab now
asks the key for every code at once, as Yubico Authenticator does,
instead of one account at a time. On a key with accounts that require a
touch, the tab no longer waits about 15 seconds per such account before
showing anything: those accounts show Read code straight away. A
Nitrokey 3 has no all-codes command, so keyroost asks for its account
list with each account's touch setting instead, with the same result.
Accounts with a period other than 30 seconds are still read one by one.
(#199)
One Refresh button. The top bar's Refresh is now the only one: the
sidebar's rescan icon, the Refresh on the Authenticator, PIV and OTP
tabs, OpenPGP's Read status, FIDO2's Reload and Load buttons and the
Molto2's Refresh slots are gone. Refresh reads again whatever is open,
down to the passkey, fingerprint or storage list. The OpenPGP card and the
fingerprint list now read themselves when you open them, like the other
tabs. The New CHUID dialog's button for a new random ID is now labeled
"New GUID". (#200)
An unplugged key is selected again when it comes back. Unplugging the
selected key used to leave whichever key took its place selected after
you plugged it back in. The app now selects the key again, unless you
picked another one meanwhile. It recognizes the key by its serial or the
name stored on it, never by a guess. (#202)
A certificate can go into a PIV slot that has no key yet, in the app too.
The app used to disable Import certificate for such a slot; it now allows
it, as the CLI and other tools do, and both say the certificate won't work
until a matching key is put in the slot. (#215)
Nitrokey is spelled Nitrokey. piv info and the app's PIV header wrote
"NitroKey"; piv --json info now reports
"applet_fingerprint": "Trussed::Nitrokey" (was Trussed::NitroKey).
(#215)
A code you touched the key for no longer vanishes at the 30-second
mark. In the app, Read code on an account that requires a touch now
works out the code for 10 seconds ahead, as Yubico Authenticator does, and
the code stays on screen until its own 30 seconds are up. When that is the
next 30 seconds, the code shows in gray with wait 4s counting down
until it starts, and its ? explains why. (#217)
A FIDO2 key with no serial number can be reset when it is the only key
plugged in. The app refused such a key outright. The app and
keyroostctl now follow one rule: with other keys connected, the reset
is refused before anything is wiped, naming the keys to unplug. (#221)
A banner asks for the touch the key is waiting for. While a job waits
for a touch, the request shows in a banner at the top of the window
instead of as small gray text in the top bar. (#221)
The idle lock also wipes secrets left in a form. A PIN, PUK, password,
passphrase, management key or seed typed into any masked box and not yet
used now starts the idle clock, and the box is emptied when it runs out.
(#223)

Fixed

Token2 OTP works over USB on PIN+ keys with firmware 2.1. On these
keys otp list, the app's On-device OTP tab and every other OTP command
failed with "unexpected status word 0x0105": the key's FIDO interface
didn't accept the way keyroost framed OTP commands. keyroost now sends
them through the CTAPHID vendor command Token2 documented, and keeps the
earlier framing for keys that don't answer it. A code that needs a touch
says "touch your key" when the key asks for one. (#155)
otp info and otp reset work over USB on the same keys. Both failed
with "unexpected status word 0x6FFB": the key rejects the short form
keyroost sent them in over USB. They now go out in the form Token2's
protocol document and reference client use; over a smart-card reader
nothing changes. (#155)
A PIV slot holding only a certificate no longer reads as holding a
key. In the app, generating a key into such a slot warned that it
"destroys the existing key", and Move key and Delete key were offered
for it. It now reads "certificate present, no key", and generating or
moving a key into it notes that the certificate stays. Retired slots now
show their certificates too. keyroostctl piv info adds ", no key" to
such a slot's line. (#174)
TrustKey keys are named TrustKey. Keys with USB vendor ID 0x311F
(registered to eWBM, whose keys now sell as TrustKey) showed as "Security
key" in 0.14.0; they now show as TrustKey, with the model from the key
(for example "TrustKey B210H"). Reported by @Ephemera42. (#195)
U2F-only keys are labeled U2F. A key whose FIDO side speaks only U2F,
such as the YubiKey NEO, was labeled FIDO2. keyroost now asks each key's
FIDO interface which protocol it speaks and labels it U2F when that is all
it does, in keyroostctl list (and list --json) and in the app, which
no longer offers a FIDO2 tab or a FIDO2 reset step for it. (#201)
Piping a secret in Windows PowerShell 5.1 works. "SEED" | keyroostctl oath add NAME --seed stdin failed there with "the seed is
not valid base32 (invalid character in input)": PowerShell 5.1 can start
piped text with an invisible byte-order mark, two of them when the
console also uses UTF-8, which keyroost took as part of the seed.
Byte-order marks at the start of stdin are now ignored for every secret
read from it. (#203)
The key's model and firmware stay in the header. After a rescan
(Refresh, or plugging a key in or out) the selected key's header dropped
back from, for example, "YubiKey 5 Series with NFC · fw 5.7.4" to
"YubiKey". It now keeps them. (#204)
Files saved by Windows tools import. An Aegis or 2FAS export or an
otpauth:// list that starts with a byte-order mark (as Windows PowerShell
5.1's Out-File -Encoding utf8 and some editors write) was refused, or
lost its first account; UTF-16 text (what PowerShell 5.1's > writes)
wasn't read at all. Both now import, in the app and with
keyroostctl … import. (#205)
Authenticator works on the Nitrokey 3. On current Nitrokey 3 firmware
every Authenticator command failed with "conditions of use not satisfied":
the key holds a reply back until it is asked for with the standard
GET RESPONSE command, and keyroost only knew the YubiKey's way of asking.
keyroost now asks each key the way it expects, so accounts can be listed,
added, read and deleted. A Nitrokey 3 asks for a touch to reset its
Authenticator app, to delete any account, and to save or read one that
requires a touch. The CLI and the app say so, and a missed touch gets a
message that says to try again and touch the key. (#208)
Authenticator codes no longer drop out while a browser is open. A
browser checking the key for smart-card certificates could slip its own
commands in between keyroost's, and the app then showed Read code on
ordinary accounts. keyroost now keeps the key to itself for the length of
each Authenticator read or change. (#208)
The Nitrokey 3's firmware version shows in the app. The key header
showed "fw 0.0.0", the placeholder the key's FIDO side reports; it now
shows the version from the key's admin application. (#208)
Device scans ask a key for its identity once. On macOS, the app asked
a Yubico key for its serial number over the FIDO connection on every
device scan, so a key that doesn't support that question (such as the
YubiKey NEO) was asked again and again, and debug logs filled with the
same two lines. keyroost now keeps the answer for as long as the key stays
plugged in, wherever that can't be mistaken for a different key.
(#211)
PIV, OpenPGP and OATH reads work on a YubiKey NEO. On macOS, reading
a YubiKey NEO's PIV, OpenPGP or OATH applet failed with "no Smart Card is
currently in the device" or "an attempt was made to end a non-existent
transaction", while ykman read the same key. The reads failed right after
keyroost asked the NEO's FIDO side for its serial number, a question it
does not answer. keyroost no longer asks it. As an added safeguard,
keyroost also waits up to three seconds for a card that has just reset and
opens the session once more, before sending anything that could change the
key. (#211)
A FIDO key that doesn't answer properly shows as FIDO2?. A key that
opened but gave no proper answer to keyroost's first question used to
show as FIDO2 and get a name read and a FIDO2 reset step; it now shows
FIDO2? and gets neither until it answers. A key keyroost can't open at
all keeps FIDO2 as before, and a failed question is asked again on the
next scan. A key reached only through a smart-card reader now shows
FIDO2? unless its FIDO applet reports FIDO2. (#215)
doctor no longer calls a Token2 FIDO key's reader a Molto2. It tagged
every reader with "TOKEN2" in its name "(Molto2)"; only a Molto2's reader
gets the tag now. (#216)
A Nitrokey 3 is called "Nitrokey 3", not "3". The CLI's target line,
the app's key header and the lists that ask which key to use showed only
"3". (#216)
Without administrator rights on Windows, "no key" errors say why.
fido info and otp --transport hid said no key was connected when a
FIDO key was plugged in. Windows only lets a program running as
administrator open a key's FIDO USB side, and the error now says so, as
list already did. (#216)
fido info --reader and fido reset --reader name the reader they use.
The target line showed the key's USB path instead. fido reset --reader
now checks the reader before asking to confirm: it gives the same error
as fido info --reader when the reader's card has no FIDO side, and on
Windows without administrator rights it says administrator rights are
needed. (#217)
fido info shows a Nitrokey 3's firmware version. It printed 0.0.0,
the number the key gives over USB; keyroost now reads the real version
from the key, as the app already did. (#217)
otp code asks for a touch through a smart-card reader too. For an
entry saved with --touch, it waited silently for the touch when it went
through the key's reader; it now says to touch the key, as it does over
USB. (#217)
The On-device OTP tab fits a narrow window. The add-entry form broke
its labels mid-word, the secret field ran past the edge, and the tab's
buttons covered its title. The buttons now move under the title when there
is no room. (#217)
Importing from a .p12 file names the key by its certificate. A file
exported by Windows showed an internal name such as te- followed by a
long code; the import now shows the certificate's name instead. (#217)
A factory reset asks for the touch that erasing OTP entries needs. On a
Token2 key the reset's OTP step waited silently for a button press and then
failed with "timed out waiting for a button press on the key". The app's
Factory reset card now says to touch the key while the reset runs, and
keyroostctl factory-reset prints a touch line over a smart-card reader
too. (#218)
Forced PIN changes no longer block unlocking. When a FIDO key
requires a PIN change, the app opens the change-PIN form instead of failing
with a PIN policy error. After the change, the key can be unlocked with the
new PIN. CLI commands that need a PIN token now ask for a PIN change first
instead of reporting a misleading new-PIN complexity error. Contributed by
@Snrat. (#220)
Idle locking clears every PIN draft. Current, new, and confirmation
PINs left in the change-PIN form are now wiped after inactivity, including
when setting a first PIN. Contributed by @Snrat. (#222)
"Always require user verification" is only offered on keys that report
it. Pressing Toggle on a key without the alwaysUv option failed with
"authenticator returned CTAP2 status 0x02". The app now leaves the row
out for such keys, and fido config always-uv says the key doesn't
support it. Contributed by @Snrat. (#224)
keyroost asks for a touch wherever the key needs one. Over USB, the
CLI and the app now say "touch your key" whenever the key reports it is
waiting for a touch. Over a smart-card connection they say so up front
where a touch is known or possible: oath code and oath delete on a
touch account, the factory reset's OATH step, OpenPGP key import and
backup restore, and Nitrokey 3 code replacement. The app's banner
follows the key, and no longer appears while unlocking. (#225)
The idle lock also closes what a typed password opened. A file read
for import, an OpenPGP key set made or read from a backup, and a .p12
key now start the idle clock, and the dialog holding them closes when it
runs out, saying why. The On-device OTP PIN boxes clear their typed
history like every other secret box. (#226)
Enterprise attestation, the minimum PIN length and a forced PIN change
are offered only on keys that support them. The app's FIDO2 Settings show those rows only
where the key lists them, and keyroostctl refuses before asking for
the PIN, as it already did for Always UV. (#227)
Each key's firmware is shown the way its maker states it. A Solo 2
showed 2.3.196 for firmware 2.964, and a Token2 PIN+ showed 2.1.2; it now
shows its release (for example "R3.4"), named from its serial number as
Token2 publishes. In keyroostctl fido info and the app's key header;
fido --json info gives firmware: null when a key doesn't say. (#228)

Security

Authenticator debug traces no longer show account names or codes.
keyroostctl --debug and the app's APDU trace showed the Authenticator's
account names (in hex) and current codes when listing or reading codes.
Those parts are now hidden, like seeds and passwords already were; the
status words stay, which is what a bug report needs. Traces get pasted
into public issues, and an account list tells which services someone
uses. (#206)
Newer elliptic-curve crates for P-256, P-384 and secp256k1. The FIDO2
PIN protocol, Token2 OTP seed encryption, OpenPGP key generation and
signing, and the PIV self-test now use the RustCrypto elliptic-curve 0.14
crates, which clears RUSTSEC-2026-0387, a timing-variability advisory for
crypto-bigint 0.5. keyroost behaves the same. Building the libraries and
the CLI from source now needs Rust 1.87 or newer (it was 1.85), because
the fixed crypto-bigint and ctutils require it; the GUI still needs 1.92.
(#212)

Full changelog: https://github.com/framefilter/keyroost/blob/v0.15.0/CHANGELOG.md