Take the 2-minute tour ×
Cryptography Stack Exchange is a question and answer site for software developers, mathematicians and others interested in cryptography. It's 100% free, no registration required.

Strange question maybe, but say I want to specify a protocol that complies with NIST SP 800-108. This lists a whole set of key based key derivation functions, with a very loose set of parameters.

How should I identify the KDF used? It seems that the KDF's in this document have neither a name, an OID nor an ASN.1 module. Are there any compliant KDF's in the world that comply to the descriptions and that do have a name/OID/ASN.1?

Or do I have to do this work myself and then test myself against the available test vectors?

share|improve this question
2  
This may not be easy to answer as even NIST (the author of 800-108) responded negatively to my request for this. In the mean time I'll use HKDF instead of 108 for my personal needs. –  owlstead Apr 26 at 15:31
add comment

1 Answer 1

up vote 2 down vote accepted

Background

When defining protocol compliant with NIST SP 800-108, you just need to pick suitable options, which work well with your protocol.

If there is a need to be compatible with a specific pre-existing protocol, you may want to take a look at NIST SP 800-135Rev1, which defines application specific key derivation functions. It is notable to recognize that some of the NIST SP 800-135Rev1 key derivation functions are valid NIST SP 800-108, with specific options selected.

The best way to identify the KDF is to actually list all the applicable options from NIST SP 800-108 KBKDF. (What of three functions to use (CTR, FB, Pipeline); order of components, what is the content of label, and if the label is terminated by zero etc.) It is likely best to illustrate the protocol specification with clear examples which show the correspondence of NIST SP 800-108 functions (or alternative show invocations of underlying MAC function) as used in key derivation.

Testing NIST SP 800-108 implementation

The recommended way to test is via CAVP (Cryptographic Algorithm Validation Program). Testing takes a long time and will be performed by a CST (Certified testing Laboratory), which costs money. But you get to choose pretty specifically what to test (from reasonably large option space) and if you pass the testing, you will get a algorithm validation certificate.

For most common options, less expensive and quicker way could be to test against pre-existing unofficial test vectors from NIST/CAVP.

If the implementation needs not only to be FIPS compliant, but also is required to have US government certification (FIPS 140-2), then algorithm certificates for the applicable algorithms are needed.

If you need a set of test vectors for your KBKDF "instantiation", one way to do it is to first ensure correctness of implementation using one of the above mechanisms, and the use that "appears to be correct" implementation to generate the official test vectors, and attach the test vectors with the protocol specification.

The way to determine order, sizes of elements etc. used in the test vectors is explained in KBKDFVS (KBKDF Validation System).

The parameters KBKDFVS uses for Feedback Mode are (extracted from KBKDFVS):

  1. Pseudo-Random Function supported by the IUT
    1. List CMAC AES 128, 192, 256
    2. CMAC TDES 2, 3
    3. HMAC SHA1, SHA224, SHA256, SHA384, SHA512
  2. L-length of the derived keying material K0 in bytes – Enter all that apply:
    1. A minimum and maximum value for full block lengths supported
    2. A minimum and maximum value for partial block lengths supported
  3. Method(s) supported to generate K:
    1. SP800-56A
    2. SP800-56B
    3. SP800-90A
    4. an Approved RNG
    5. N/A – Out of the scope of the algorithm implementation
  4. Does the IUT support the counter being used as an input? If yes, what is the length of the binary representation of the counter i.
    1. 8 bits.
    2. 16 bits.
    3. 24 bits.
    4. 32 bits.
  5. Does the IUT support the counter coming before and/or after the iteration variable and/or after the fixed input data?

The information for other modes (i.e.) counter and double pipeline is similar. This information is not sufficient to encode all that can be known about the KDF function, but is already a good start.

Finally addressing your questions

  • Identify the KDF by specifying options from NIST SP 800-108. You should also have examples (and possibly test vectors), which allow correct implementation.
  • The name is of NIST SP 800-108 and the name (or section) of the mode used.
  • I'm also not aware of OID/ASN.1 applicable. (But maybe somebody has defined something for their implementations of NIST SP 800-108. There are no OID/ASN.1 which covers all the possible options of NIST SP 800-108.)
  • If you need OID or ASN.1 module for your protocol, it is likely easiest for now to define your own.
  • Implementations of SP 800-108 likely can be tested using CAVP or unofficial CAVP test vectors.
  • KBKDFVS can be good example to see how NIST specifies the parameters in context of test vectors. Maybe you can use similar approach and terminology for the protocol?

HKDF

Side Note: Depending on how you use HKDF, you are often using either NIST SP 800-108 or NIST SP 800-56C: HKDF "expand only" can be considered to be a variant of NIST SP 800-108 KBKDF. NIST SP 800-56C is basically attempt to define full HKDF in NIST terminology (and add AES-CMAC).

share|improve this answer
    
Ah, I hadn't noticed 56C before, I'm off studying. I'm aware of the test vectors for 800-108, and I've tested all of them against my implementations (that I've created for Bouncy Castle). What I'm unaware of is a method of specifying a unique KDF. I mean, order, separators, hash function, paragraph of 800-108 (none has an official name), salt size... –  owlstead Apr 26 at 19:08
    
@owlstead: Have you taken a look at the scheme for describing the contents of test vectors, given in KBKDFVS? It describes a lot about the usage of the KDF. I improved my answer to cover that. –  user4982 Apr 27 at 13:17
    
Of course, as I had to parse them, I wasn't going 40 MiB of test vectors by hand :), but thanks for including them in the answer. It's probably the best way of specifying the hash, although I had to fix some misconceptions of the test vectors before my implementation turned up with just "green flags". So the test vectors aren't that well defined either... –  owlstead Apr 27 at 13:36
    
I'll leave this question running for a while if you don't mind. If I don't get a direct answer I'll accept this one until one turns up later. Hopefully. Some day :) –  owlstead Apr 28 at 7:50
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.