Copyright © 2013-2020 FIDO Alliance All Rights Reserved.
This document helps support the FIDO Authenticator Security Certification program. This list does not in any way alter the protocol specifications provided in other FIDO Authenticator documents, so the presence or absence of an algorithm in this list does not suggest that this algorithm is or is not allowed within any FIDO protocol. For certified FIDO Authenticators, there are various requirements that limit “internal” algorithms, those that are not explicitly specified within the FIDO Authenticator protocol. Additionally, the procedure for determining the “Overall Authenticator Claimed Cryptographic Strength” involves locating the security level for each algorithm used by the FIDO Authenticator within this document; this procedure applies to all cryptographic algorithms used by the FIDO Authenticator.
This section describes the status of this document at the time of its publication. Other documents may supersede this document. The most recent version of this document can be found on the FIDO Alliance Website at https://fidoalliance.org.
This document was published by the FIDO Alliance as a Final Requirements Document. If you wish to make comments regarding this document, please Contact Us. All comments are welcome.
No rights are granted to prepare derivative works of this document. Entities seeking permission to reproduce portions of this document for other uses must contact the FIDO Alliance to determine whether an appropriate license for such use is available.
Implementation of certain elements of this Requirements Document may require licenses under third party intellectual property rights, including without limitation, patent rights. The FIDO Alliance, Inc. and its Members and any other contributors to the Requirements Document are not, and shall not be held, responsible in any manner for identifying or failing to identify any or all such third party intellectual property rights.
THIS FIDO ALLIANCE REQUIREMENTS DOCUMENT IS PROVIDED “AS IS” AND WITHOUT ANY WARRANTY OF ANY KIND, INCLUDING, WITHOUT LIMITATION, ANY EXPRESS OR IMPLIED WARRANTY OF NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in [RFC2119].
If a vendor wants to use a cryptographic security function for an internal use that requires an Allowed algorithm, or to claim a non-zero security strength, then the vendor / lab shall provide a written argument that it:
The stated security level identifies the expected number of computations that a storage-constrained attacker (who has access to no more than 2^80 bytes of storage) shall expend in order to compromise the security of the cryptographic security function, under the currently best known attack that can be conducted under this storage constraint. This has been extracted from the currently best known relevant attacks against each cryptographic primitive, and is expected to shift over time as attacks improve.
At the time the document is published, there are not yet any standardized (NIST Project https://csrc.nist.gov/projects/post-quantum-cryptography or ISO/IEC JTC 1/SC 27/WG 2 SD8 on Post-Quantum Cryptography) quantum-safe cryptographic algorithms for asymmetric algorithms (e.g. signature, key protection based on RSA, and anonymous attestation). It is also not yet clear if the key size should (or should not) be increased for symmetric algorithms.
If the security level stated is n (where the security level is here defined with a classical computing power, and does not take into account quantum cryptanalysis), then the expected number of computations is less than the expected number of computations required to guess an (n+1)-bit random binary string, and not less than the number of computations required to guess an n bit random binary string (i.e., on average, the number of computations required is less than 2^n computations and greater than or equal to 2^(n-1) computations).
Provide confidentiality, up to the stated security level.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
AES-128 | [FIPS197], [ISOIEC-18033-3] | 128 |
AES-192 | [FIPS197], [ISOIEC-18033-3] | 192 |
AES-256 | [FIPS197], [ISOIEC-18033-3] | 256 |
Provide pre-image resistance, 2nd pre-image resistance, and collision resistance.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
SHA-256 | [FIPS180-4], [ISOIEC-10118-3] | 128 |
SHA-384 | [FIPS180-4], [ISOIEC-10118-3] | 192 |
SHA-512 | [FIPS180-4], [ISOIEC-10118-3] | 256 |
SHA-512/t, 256 ≤ t < 512 | [FIPS180-4], [ISOIEC-10118-3] | t/2 |
SHA3-256 | [FIPS202], [ISOIEC-10118-3] | 128 |
SHA3-384 | [FIPS202], [ISOIEC-10118-3] | 192 |
SHA3-512 | [FIPS202], [ISOIEC-10118-3] | 256 |
Provide data authentication.
It is not uncommon to truncate the result of a MAC, but to be resistant, the final length should be at least 96 bits (see SOGIS Note 14-MACTruncation96 in [SOGISCrypto]). The Security Level cannot exceed the final length.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
HMAC | [FIPS198-1] | Minimum of the length of the output of the hash used^{[2]}, one-half of the number of bits in the hash state^{[3]}, or the number of bits in the HMAC key. |
CMAC | [SP800-38B] | Equal to the minimum of the strength of the underlying cipher and the length of the output MAC. |
GMAC | [SP800-38D] | Equal to the minimum of the strength of the underlying cipher and the length of the output MAC. |
[2]Both due to the obvious guessing attack, and covers the case where the supplied key is hashed for the HMAC.
[3]Based on a birthday attack; a collision of the final state can lead to an existential forgery of longer messages with the same prefix.
The underlying cryptographic algorithms used in HMAC, CMAC or GMAC need to be 112-bit cryptography strength at least and part of the list of FIDO allowed cryptographic algorithms.
SOGIS gives some additional recommendations for the use of GMAC, including IV (length must be equal to 96 bits) or MAC length (must be one of the values 96, 104, 112, 120, and 128 bits). See [SOGISCrypto] for the complete list of recommendations.
Provide confidentiality and data authentication.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
Key Wrapping | [SP800-38F] | Equal to the strength of the underlying cipher. |
GCM Mode, with length 96 bit or larger IVs. For any given key, the IV length must be fixed. | [SP800-38D] | Equal to the strength of the underlying cipher. |
RSA OAEP | [RFC3447]. Key generation must be according to [FIPS186-4]. | Depends on the parameter size: according to NIST, 112 bits for RSA 2048 and 128 bits for RSA 3072 |
CCM Mode | [SP800-38C] | Equal to the strength of the underlying cipher. |
Encrypt-then-HMAC^{[4]} | Encryption specification depends on the cipher selected. HMAC specification [FIPS198-1] | The minimum of the strength of the cipher and the HMAC. |
Encrypt-then-CMAC^{[5]} | Encryption specification depends on the cipher selected. CMAC specification [SP800-38B] | The minimum of the strength of the cipher and the CMAC. |
[4]The cipher and HMAC shall use independent keys, and the information HMACed shall include any IV / Nonce / Counter (if sent/stored), and, if the message size varies, the length of the message; when present, this message length shall reside prior to any variable length message components.
[5]The cipher and CMAC shall use independent keys, and the information CMACed shall include any IV / Nonce / Counter (if sent/stored).
The underlying cryptographic algorithms used in the above table need to be 112-bit cryptography strength at least and part of the list of FIDO allowed cryptographic algorithms.
Evidence that the requirements are met could be given by providing a proof that the implementation uses the underlying platform certified RNG/RBG through Common Criteria, FIPS 140-3 (issued on March 22nd 2019 or after) or an equivalent evaluation scheme against the listed standards, or by having a FIDO approved lab conducting an evaluation of the RNG/RBG implementation against the standards listed below. In other words, the following standards define the metrics required to assess the quality of the RNG implementation.
FIPS 140-3 became effective on September 22, 2019.
If the designer is interested in retaining the security of an (EC)DSA private key in the event of an entropy source failure or Deterministic Random Number Generator state compromise, then RFC6979-like properties can be obtained by providing the hash of the message being signed and the private key in use to the Deterministic Random Number Generator in a secure fashion (e.g., via the SP800-90A additional input parameter). Additional parameters (e.g., the KeyID / Key Handle, if it was randomly generated) may also be used to increase resistance to attack in certain scenarios.
In the last version of FIPS 140-2, IG7.18 gives a clear rule on the seeding sources for allowed DRBGs: Entropy Estimation and Compliance with [SP800-90b] is mandatory.
The (physical) random number generator shall meet the requirements specified in:
If PTG.2 is used, an application-specific post processing may additionally be required to prevent any bias in the output function.
For instance, these requirements are met if a certified hardware platform is used (e.g. according to Global Platform TEE Protection Profile or Eurosmart Security IC Platform Protection Profile) and the Security Target contains Extended Component FCS_RNG.1 including at least one of the allowed classes PTG.2, or PTG.3.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
Source RBG is DRBG with access to Live Entropy Source or it is an NRBG. | [SP800-90C], section 6 | Any security strength. |
We consider this a physical RNG if at least as much entropy is added into the RNG as is retrieved per request.
It is uncommon for the DRBGs in FIPS modules to meet these requirements, unless their design anticipates one of the SP800-90C NRBG designs.
The security strength (in bits) of an allowed physical/true random number generator is equivalent to the size (in bits) of the random bytes retrieved from it.
Provide computational indistinguishability from an ideal random sequence, cycle resistance, non-destructive reseeding, insensitivity of a seeded generator to seed source failure or compromise, backtracking resistance. Ideally, the ability to provide additional input, and ability to recover from a compromised internal state.
The (deterministic) random number generator shall meet the requirements specified in:
Algorithm | Specified in | Security Level (bits) |
---|---|---|
HMAC_DRBG | [SP800-90ar1], Revision 1, section 10.1.2 | The instantiated security level, as defined in [SP800-90ar1]. |
CTR_DRBG | [SP800-90ar1], Revision 1, section 10.2.1 | The instantiated security level, as defined in [SP800-90ar1]. |
HASH_DRBG | [SP800-90ar1], Revision 1, section 10.1.1 | The instantiated security level, as defined in [SP800-90ar1]. |
We consider this a deterministic RNG if less entropy is added into the RNG than is retrieved.
The last version of FIPS 140-2 [FIPS140-2] IG 7.18 requires transition to full compliance with [SP800-90b] for seed for DRBG.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
Diffie-Hellmann (DH) 2048-bit key | [SP800-56Ar3], [ISOIEC-11770-3] | >=112 |
EC-DH on P-256 or BrainpoolP256r1 | [SP800-56Ar3], [ISOIEC-11770-3] | >=128 |
Algorithm | Specified in | Security Level (bits) |
---|---|---|
KDF in counter mode | [SP800-108] | min(Bit length of key derivation key Ki used as input, Security level of PRF) |
KDF in feedback mode | [SP800-108] | min(Bit length of key derivation key Ki used as input, Security level of PRF) |
KDF in double pipeline iteration mode | [SP800-108] | min(Bit length of key derivation key Ki used as input, Security level of PRF) |
HKDF | [SP800-56cr1], [RFC5869] | min(Bit length of key derivation key Ki used as input, Security level of HMAC) |
Where PRF denotes an acceptable pseudorandom function (based on HMAC or CMAC), and key-derivation key denotes an acceptable pre-shared cryptographic key (with key length at least 112 bits), as defined in [SP800-108].
The underlying cryptographic algorithms used in the above table need to be 112-bit cryptography strength at least and part of the list of FIDO allowed cryptographic algorithms.
The KDF used in CTAP for PIN protocol version 1 is slightly different from [SP800-56cr1], but is acceptable from a security point of view.
Provide data authentication, and non-repudiation.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
ECDSA on NIST P-256 | [ECDSA-ANSI], [FIPS186-4], [ISOIEC-14888-3] | 128 |
ECDSA on BrainpoolP256r1 | [RFC5639] | 128 |
ECDSA on secp256k1 | [ECDSA-ANSI], [FIPS186-4], [[!Certicom SEC 2]] | 126^{[7]} |
2048-bit RSA PSS | [FIPS186-4],[ISOIEC-9796-2] | 112 |
1024*n-bit RSA PKCS v1.5 (n=2,3,4) | [FIPS186-4],[ISOIEC-9796-2] | 112 |
1024*n-bit RSA PKCS v2.1 (PSS) (n=2,3,4) | [FIPS186-4],[ISOIEC-9796-2] | 112 |
SM2 digital signatures (SM2 part 2) using the SM3 hash on the SM2 curve specified by OSCCA. | SM2 [ISOIEC-14888-3] and SM3 [ISOIEC-10118-3] | 128 |
Ed25519 | EDDSA [RFC8032] | 128^{[8]} |
[7] Based on an attack using Pollard rho on the equivalence classes defined by the curve’s easily computable endomorphism.
[8] Based on the difficulty of performing discrete logs on the group defined by the recommended curve parameters.
NIST [SP800-131Ar2] disallows the use of PKCS#1 version 1.5 after December 31, 2023. SOGIS [SOGISCrypto] only allow it for legacy reasons (until end 2022).
Provide anonymous attestation.
In most cases, the limiting factor was the difficulty of performing the discrete log calculation within the embedding field.
The security level values here were taken from NIST guidance. This NIST guidance is based on conducting the discrete log calculation within prime ordered fields; the structure of the fields here is richer, and this structure could possibly allow for a more advanced discrete log approach that could be considerably faster. Currently, the best known algorithms in both cases have the same asymptotic complexity (Lq [1⁄3]), but without extensive testing, it isn’t clear how the number of computations compares.
In addition, the NIST guidance does not allow for security levels other than a few specific proscribed values: if the number of bits required to represent the order of the embedding field is between 3072 and 7679, the security level is reported as 128 bits. Similarly, if the number of bits required to represent the order of the embedding field is between 2048 and 3071, the security strength is reported as 112 bits.
Algorithm | Specified in | Security Level (bits) |
---|---|---|
ED256-2 | [FIDOEcdaaAlgorithm], section Object Formats and Algorithm Details, [DevScoDah2007] | 112 |
ED512 | [FIDOEcdaaAlgorithm], section Object Formats and Algorithm Details, [ISO15946-5] | 128 |
ED638 | [FIDOEcdaaAlgorithm], section Object Formats and Algorithm Details, [TPMv2-Part4] | 128 |