openssh/PROTOCOL.u2f
markus@openbsd.org b556cc3cbf upstream: remove extra layer for ed25519 signature; ok djm@
OpenBSD-Commit-ID: 7672d9d0278b4bf656a12d3aab0c0bfe92a8ae47
2019-11-13 08:54:09 +11:00

237 lines
8.2 KiB
Plaintext

This document describes OpenSSH's support for U2F/FIDO security keys.
Background
----------
U2F is an open standard for two-factor authentication hardware, widely
used for user authentication to websites. U2F tokens are ubiquitous,
available from a number of manufacturers and are currently by far the
cheapest way for users to achieve hardware-backed credential storage.
The U2F protocol however cannot be trivially used as an SSH protocol key
type as both the inputs to the signature operation and the resultant
signature differ from those specified for SSH. For similar reasons,
integration of U2F devices cannot be achieved via the PKCS#11 API.
U2F also offers a number of features that are attractive in the context
of SSH authentication. They can be configured to require indication
of "user presence" for each signature operation (typically achieved
by requiring the user touch the key). They also offer an attestation
mechanism at key enrollment time that can be used to prove that a
given key is backed by hardware. Finally the signature format includes
a monotonic signature counter that can be used (at scale) to detect
concurrent use of a private key, should it be extracted from hardware.
U2F private keys are generated through an enrollment operation,
which takes an application ID - a URL-like string, typically "ssh:"
in this case, but a HTTP origin for the case of web authentication,
and a challenge string (typically randomly generated). The enrollment
operation returns a public key, a key handle that must be used to invoke
the hardware-backed private key, some flags and signed attestation
information that may be used to verify that a private key is hosted on a
particular hardware instance.
It is common for U2F hardware to derive private keys from the key handle
in conjunction with a small per-device secret that is unique to the
hardware, thus requiring little on-device storage for an effectively
unlimited number of supported keys. This drives the requirement that
the key handle be supplied for each signature operation. U2F tokens
primarily use ECDSA signatures in the NIST-P256 field.
SSH U2F Key formats
-------------------
OpenSSH integrates U2F as a new key and corresponding certificate type:
sk-ecdsa-sha2-nistp256@openssh.com
sk-ecdsa-sha2-nistp256-cert-v01@openssh.com
These key types are supported only for user authentication with the
"publickey" method. They are not used for host-based user authentication
or server host key authentication.
While each uses ecdsa-sha256-nistp256 as the underlying signature primitive,
keys require extra information in the public and private keys, and in
the signature object itself. As such they cannot be made compatible with
the existing ecdsa-sha2-nistp* key types.
The format of a sk-ecdsa-sha2-nistp256@openssh.com public key is:
string "sk-ecdsa-sha2-nistp256@openssh.com"
ec_point Q
string application (user-specified, but typically "ssh:")
The corresponding private key contains:
string "sk-ecdsa-sha2-nistp256@openssh.com"
ec_point Q
string application (user-specified, but typically "ssh:")
string key_handle
uint32 flags
string reserved
The certificate form of a SSH U2F key appends the usual certificate
information to the public key:
string "sk-ecdsa-sha2-nistp256-cert-v01@openssh.com"
string nonce
ec_point Q
string application
uint64 serial
uint32 type
string key id
string valid principals
uint64 valid after
uint64 valid before
string critical options
string extensions
string reserved
string signature key
string signature
During key generation, the hardware also returns attestation information
that may be used to cryptographically prove that a given key is
hardware-backed. Unfortunately, the protocol required for this proof is
not privacy-preserving and may be used to identify U2F tokens with at
least manufacturer and batch number granularity. For this reason, we
choose not to include this information in the public key or save it by
default.
Attestation information is very useful however in an organisational
context, where it may be used by a CA as part of certificate
issuance. In this case, exposure to the CA of hardware identity is
desirable. To support this case, OpenSSH optionally allows retaining the
attestation information at the time of key generation. It will take the
following format:
string "sk-attest-v00"
uint32 version (1 for U2F, 2 for FIDO2 in future)
string attestation certificate
string enrollment signature
SSH U2F signatures
------------------
In addition to the message to be signed, the U2F signature operation
requires a few additional parameters:
byte control bits (e.g. "user presence required" flag)
byte[32] SHA256(message)
byte[32] SHA256(application)
byte key_handle length
byte[] key_handle
This signature is signed over a blob that consists of:
byte[32] SHA256(application)
byte flags (including "user present", extensions present)
uint32 counter
byte[] extensions
byte[32] SHA256(message)
The signature returned from U2F hardware takes the following format:
byte flags (including "user present")
uint32 counter
byte[32] ecdsa_signature (in X9.62 format).
For use in the SSH protocol, we wish to avoid server-side parsing of ASN.1
format data in the pre-authentication attack surface. Therefore, the
signature format used on the wire in SSH2_USERAUTH_REQUEST packets will
be reformatted slightly and the ecdsa_signature_blob value has the encoding:
mpint r
mpint s
byte flags
uint32 counter
Where 'r' and 's' are extracted by the client or token middleware from the
ecdsa_signature field returned from the hardware.
For Ed25519 keys the signature is encoded as:
string "sk-ssh-ed25519@openssh.com"
string signature
byte flags
uint32 counter
ssh-agent protocol extensions
-----------------------------
ssh-agent requires a protocol extension to support U2F keys. At
present the closest analogue to Security Keys in ssh-agent are PKCS#11
tokens, insofar as they require a middleware library to communicate with
the device that holds the keys. Unfortunately, the protocol message used
to add PKCS#11 keys to ssh-agent does not include any way to send the
key handle to the agent as U2F keys require.
To avoid this, without having to add wholly new messages to the agent
protocol, we will use the existing SSH2_AGENTC_ADD_ID_CONSTRAINED message
with a new key constraint extension to encode a path to the middleware
library for the key. The format of this constraint extension would be:
byte SSH_AGENT_CONSTRAIN_EXTENSION
string sk@openssh.com
string middleware path
This constraint-based approach does not present any compatibility
problems.
OpenSSH integration
-------------------
U2F tokens may be attached via a number of means, including USB and NFC.
The USB interface is standardised around a HID protocol, but we want to
be able to support other transports as well as dummy implementations for
regress testing. For this reason, OpenSSH shall perform all U2F operations
via a dynamically-loaded middleware library.
The middleware library need only expose a handful of functions:
/* Flags */
#define SSH_SK_USER_PRESENCE_REQD 0x01
/* Algs */
#define SSH_SK_ECDSA 0x00
#define SSH_SK_ED25519 0x01
struct sk_enroll_response {
uint8_t *public_key;
size_t public_key_len;
uint8_t *key_handle;
size_t key_handle_len;
uint8_t *signature;
size_t signature_len;
uint8_t *attestation_cert;
size_t attestation_cert_len;
};
struct sk_sign_response {
uint8_t flags;
uint32_t counter;
uint8_t *sig_r;
size_t sig_r_len;
uint8_t *sig_s;
size_t sig_s_len;
};
/* Return the version of the middleware API */
uint32_t sk_api_version(void);
/* Enroll a U2F key (private key generation) */
int sk_enroll(int alg, const uint8_t *challenge, size_t challenge_len,
const char *application, uint8_t flags,
struct sk_enroll_response **enroll_response);
/* Sign a challenge */
int sk_sign(int alg, const uint8_t *message, size_t message_len,
const char *application,
const uint8_t *key_handle, size_t key_handle_len,
uint8_t flags, struct sk_sign_response **sign_response);
In OpenSSH, these will be invoked by generalising the existing
ssh-pkcs11-helper mechanism to provide containment of the middleware from
ssh-agent.