首页 > 代码库 > HMAC

HMAC

2024-08-11 15:29:42   220人阅读

介绍文档及C实现

http://www.ietf.org/rfc/rfc2104.txt

 

  
  
  
  
  
  
 Network Working Group H. Krawczyk
 Request for Comments: 2104 IBM
 Category: Informational M. Bellare
 UCSD
 R. Canetti
 IBM
 February 1997
  
  
 HMAC: Keyed-Hashing for Message Authentication
  
 Status of This Memo
  
 This memo provides information for the Internet community. This memo
 does not specify an Internet standard of any kind. Distribution of
 this memo is unlimited.
  
 Abstract
  
 This document describes HMAC, a mechanism for message authentication
 using cryptographic hash functions. HMAC can be used with any
 iterative cryptographic hash function, e.g., MD5, SHA-1, in
 combination with a secret shared key. The cryptographic strength of
 HMAC depends on the properties of the underlying hash function.
  
 1. Introduction
  
 Providing a way to check the integrity of information transmitted
 over or stored in an unreliable medium is a prime necessity in the
 world of open computing and communications. Mechanisms that provide
 such integrity check based on a secret key are usually called
 "message authentication codes" (MAC). Typically, message
 authentication codes are used between two parties that share a secret
 key in order to validate information transmitted between these
 parties. In this document we present such a MAC mechanism based on
 cryptographic hash functions. This mechanism, called HMAC, is based
 on work by the authors [BCK1] where the construction is presented and
 cryptographically analyzed. We refer to that work for the details on
 the rationale and security analysis of HMAC, and its comparison to
 other keyed-hash methods.
  
  
  
  
  
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 1]
 
 RFC 2104 HMAC February 1997
  
  
 HMAC can be used in combination with any iterated cryptographic hash
 function. MD5 and SHA-1 are examples of such hash functions. HMAC
 also uses a secret key for calculation and verification of the
 message authentication values. The main goals behind this
 construction are
  
 * To use, without modifications, available hash functions.
 In particular, hash functions that perform well in software,
 and for which code is freely and widely available.
  
 * To preserve the original performance of the hash function without
 incurring a significant degradation.
  
 * To use and handle keys in a simple way.
  
 * To have a well understood cryptographic analysis of the strength of
 the authentication mechanism based on reasonable assumptions on the
 underlying hash function.
  
 * To allow for easy replaceability of the underlying hash function in
 case that faster or more secure hash functions are found or
 required.
  
 This document specifies HMAC using a generic cryptographic hash
 function (denoted by H). Specific instantiations of HMAC need to
 define a particular hash function. Current candidates for such hash
 functions include SHA-1 [SHA], MD5 [MD5], RIPEMD-128/160 [RIPEMD].
 These different realizations of HMAC will be denoted by HMAC-SHA1,
 HMAC-MD5, HMAC-RIPEMD, etc.
  
 Note: To the date of writing of this document MD5 and SHA-1 are the
 most widely used cryptographic hash functions. MD5 has been recently
 shown to be vulnerable to collision search attacks [Dobb]. This
 attack and other currently known weaknesses of MD5 do not compromise
 the use of MD5 within HMAC as specified in this document (see
 [Dobb]); however, SHA-1 appears to be a cryptographically stronger
 function. To this date, MD5 can be considered for use in HMAC for
 applications where the superior performance of MD5 is critical. In
 any case, implementers and users need to be aware of possible
 cryptanalytic developments regarding any of these cryptographic hash
 functions, and the eventual need to replace the underlying hash
 function. (See section 6 for more information on the security of
 HMAC.)
  
  
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 2]
 
 RFC 2104 HMAC February 1997
  
  
 2. Definition of HMAC
  
 The definition of HMAC requires a cryptographic hash function, which
 we denote by H, and a secret key K. We assume H to be a cryptographic
 hash function where data is hashed by iterating a basic compression
 function on blocks of data. We denote by B the byte-length of such
 blocks (B=64 for all the above mentioned examples of hash functions),
 and by L the byte-length of hash outputs (L=16 for MD5, L=20 for
 SHA-1). The authentication key K can be of any length up to B, the
 block length of the hash function. Applications that use keys longer
 than B bytes will first hash the key using H and then use the
 resultant L byte string as the actual key to HMAC. In any case the
 minimal recommended length for K is L bytes (as the hash output
 length). See section 3 for more information on keys.
  
 We define two fixed and different strings ipad and opad as follows
 (the ‘i‘ and ‘o‘ are mnemonics for inner and outer):
  
 ipad = the byte 0x36 repeated B times
 opad = the byte 0x5C repeated B times.
  
 To compute HMAC over the data `text‘ we perform
  
 H(K XOR opad, H(K XOR ipad, text))
  
 Namely,
  
 (1) append zeros to the end of K to create a B byte string
 (e.g., if K is of length 20 bytes and B=64, then K will be
 appended with 44 zero bytes 0x00)
 (2) XOR (bitwise exclusive-OR) the B byte string computed in step
 (1) with ipad
 (3) append the stream of data ‘text‘ to the B byte string resulting
 from step (2)
 (4) apply H to the stream generated in step (3)
 (5) XOR (bitwise exclusive-OR) the B byte string computed in
 step (1) with opad
 (6) append the H result from step (4) to the B byte string
 resulting from step (5)
 (7) apply H to the stream generated in step (6) and output
 the result
  
 For illustration purposes, sample code based on MD5 is provided as an
 appendix.
  
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 3]
 
 RFC 2104 HMAC February 1997
  
  
 3. Keys
  
 The key for HMAC can be of any length (keys longer than B bytes are
 first hashed using H). However, less than L bytes is strongly
 discouraged as it would decrease the security strength of the
 function. Keys longer than L bytes are acceptable but the extra
 length would not significantly increase the function strength. (A
 longer key may be advisable if the randomness of the key is
 considered weak.)
  
 Keys need to be chosen at random (or using a cryptographically strong
 pseudo-random generator seeded with a random seed), and periodically
 refreshed. (Current attacks do not indicate a specific recommended
 frequency for key changes as these attacks are practically
 infeasible. However, periodic key refreshment is a fundamental
 security practice that helps against potential weaknesses of the
 function and keys, and limits the damage of an exposed key.)
  
 4. Implementation Note
  
 HMAC is defined in such a way that the underlying hash function H can
 be used with no modification to its code. In particular, it uses the
 function H with the pre-defined initial value IV (a fixed value
 specified by each iterative hash function to initialize its
 compression function). However, if desired, a performance
 improvement can be achieved at the cost of (possibly) modifying the
 code of H to support variable IVs.
  
 The idea is that the intermediate results of the compression function
 on the B-byte blocks (K XOR ipad) and (K XOR opad) can be precomputed
 only once at the time of generation of the key K, or before its first
 use. These intermediate results are stored and then used to
 initialize the IV of H each time that a message needs to be
 authenticated. This method saves, for each authenticated message,
 the application of the compression function of H on two B-byte blocks
 (i.e., on (K XOR ipad) and (K XOR opad)). Such a savings may be
 significant when authenticating short streams of data. We stress
 that the stored intermediate values need to be treated and protected
 the same as secret keys.
  
 Choosing to implement HMAC in the above way is a decision of the
 local implementation and has no effect on inter-operability.
  
  
  
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 4]
 
 RFC 2104 HMAC February 1997
  
  
 5. Truncated output
  
 A well-known practice with message authentication codes is to
 truncate the output of the MAC and output only part of the bits
 (e.g., [MM, ANSI]). Preneel and van Oorschot [PV] show some
 analytical advantages of truncating the output of hash-based MAC
 functions. The results in this area are not absolute as for the
 overall security advantages of truncation. It has advantages (less
 information on the hash result available to an attacker) and
 disadvantages (less bits to predict for the attacker). Applications
 of HMAC can choose to truncate the output of HMAC by outputting the t
 leftmost bits of the HMAC computation for some parameter t (namely,
 the computation is carried in the normal way as defined in section 2
 above but the end result is truncated to t bits). We recommend that
 the output length t be not less than half the length of the hash
 output (to match the birthday attack bound) and not less than 80 bits
 (a suitable lower bound on the number of bits that need to be
 predicted by an attacker). We propose denoting a realization of HMAC
 that uses a hash function H with t bits of output as HMAC-H-t. For
 example, HMAC-SHA1-80 denotes HMAC computed using the SHA-1 function
 and with the output truncated to 80 bits. (If the parameter t is not
 specified, e.g. HMAC-MD5, then it is assumed that all the bits of the
 hash are output.)
  
 6. Security
  
 The security of the message authentication mechanism presented here
 depends on cryptographic properties of the hash function H: the
 resistance to collision finding (limited to the case where the
 initial value is secret and random, and where the output of the
 function is not explicitly available to the attacker), and the
 message authentication property of the compression function of H when
 applied to single blocks (in HMAC these blocks are partially unknown
 to an attacker as they contain the result of the inner H computation
 and, in particular, cannot be fully chosen by the attacker).
  
 These properties, and actually stronger ones, are commonly assumed
 for hash functions of the kind used with HMAC. In particular, a hash
 function for which the above properties do not hold would become
 unsuitable for most (probably, all) cryptographic applications,
 including alternative message authentication schemes based on such
 functions. (For a complete analysis and rationale of the HMAC
 function the reader is referred to [BCK1].)
  
  
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 5]
 
 RFC 2104 HMAC February 1997
  
  
 Given the limited confidence gained so far as for the cryptographic
 strength of candidate hash functions, it is important to observe the
 following two properties of the HMAC construction and its secure use
 for message authentication:
  
 1. The construction is independent of the details of the particular
 hash function H in use and then the latter can be replaced by any
 other secure (iterative) cryptographic hash function.
  
 2. Message authentication, as opposed to encryption, has a
 "transient" effect. A published breaking of a message authentication
 scheme would lead to the replacement of that scheme, but would have
 no adversarial effect on information authenticated in the past. This
 is in sharp contrast with encryption, where information encrypted
 today may suffer from exposure in the future if, and when, the
 encryption algorithm is broken.
  
 The strongest attack known against HMAC is based on the frequency of
 collisions for the hash function H ("birthday attack") [PV,BCK2], and
 is totally impractical for minimally reasonable hash functions.
  
 As an example, if we consider a hash function like MD5 where the
 output length equals L=16 bytes (128 bits) the attacker needs to
 acquire the correct message authentication tags computed (with the
 _same_ secret key K!) on about 2**64 known plaintexts. This would
 require the processing of at least 2**64 blocks under H, an
 impossible task in any realistic scenario (for a block length of 64
 bytes this would take 250,000 years in a continuous 1Gbps link, and
 without changing the secret key K during all this time). This attack
 could become realistic only if serious flaws in the collision
 behavior of the function H are discovered (e.g. collisions found
 after 2**30 messages). Such a discovery would determine the immediate
 replacement of the function H (the effects of such failure would be
 far more severe for the traditional uses of H in the context of
 digital signatures, public key certificates, etc.).
  
 Note: this attack needs to be strongly contrasted with regular
 collision attacks on cryptographic hash functions where no secret key
 is involved and where 2**64 off-line parallelizable (!) operations
 suffice to find collisions. The latter attack is approaching
 feasibility [VW] while the birthday attack on HMAC is totally
 impractical. (In the above examples, if one uses a hash function
 with, say, 160 bit of output then 2**64 should be replaced by 2**80.)
  
  
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 6]
 
 RFC 2104 HMAC February 1997
  
  
 A correct implementation of the above construction, the choice of
 random (or cryptographically pseudorandom) keys, a secure key
 exchange mechanism, frequent key refreshments, and good secrecy
 protection of keys are all essential ingredients for the security of
 the integrity verification mechanism provided by HMAC.
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 7]
 
 RFC 2104 HMAC February 1997
  
  
 Appendix -- Sample Code
  
 For the sake of illustration we provide the following sample code for
 the implementation of HMAC-MD5 as well as some corresponding test
 vectors (the code is based on MD5 code as described in [MD5]).
  
 /*
 ** Function: hmac_md5
 */
  
 void
 hmac_md5(text, text_len, key, key_len, digest)
 unsigned char* text; /* pointer to data stream */
 int text_len; /* length of data stream */
 unsigned char* key; /* pointer to authentication key */
 int key_len; /* length of authentication key */
 caddr_t digest; /* caller digest to be filled in */
  
 {
 MD5_CTX context;
 unsigned char k_ipad[65]; /* inner padding -
 * key XORd with ipad
 */
 unsigned char k_opad[65]; /* outer padding -
 * key XORd with opad
 */
 unsigned char tk[16];
 int i;
 /* if key is longer than 64 bytes reset it to key=MD5(key) */
 if (key_len > 64) {
  
 MD5_CTX tctx;
  
 MD5Init(&tctx);
 MD5Update(&tctx, key, key_len);
 MD5Final(tk, &tctx);
  
 key = tk;
 key_len = 16;
 }
  
 /*
 * the HMAC_MD5 transform looks like:
 *
 * MD5(K XOR opad, MD5(K XOR ipad, text))
 *
 * where K is an n byte key
 * ipad is the byte 0x36 repeated 64 times
  
  
  
 Krawczyk, et. al. Informational [Page 8]
 
 RFC 2104 HMAC February 1997
  
  
 * opad is the byte 0x5c repeated 64 times
 * and text is the data being protected
 */
  
 /* start out by storing key in pads */
 bzero( k_ipad, sizeof k_ipad);
 bzero( k_opad, sizeof k_opad);
 bcopy( key, k_ipad, key_len);
 bcopy( key, k_opad, key_len);
  
 /* XOR key with ipad and opad values */
 for (i=0; i<64; i++) {
 k_ipad[i] ^= 0x36;
 k_opad[i] ^= 0x5c;
 }
 /*
 * perform inner MD5
 */
 MD5Init(&context); /* init context for 1st
 * pass */
 MD5Update(&context, k_ipad, 64) /* start with inner pad */
 MD5Update(&context, text, text_len); /* then text of datagram */
 MD5Final(digest, &context); /* finish up 1st pass */
 /*
 * perform outer MD5
 */
 MD5Init(&context); /* init context for 2nd
 * pass */
 MD5Update(&context, k_opad, 64); /* start with outer pad */
 MD5Update(&context, digest, 16); /* then results of 1st
 * hash */
 MD5Final(digest, &context); /* finish up 2nd pass */
 }
  
 Test Vectors (Trailing ‘\0‘ of a character string not included in test):
  
 key = 0x0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b0b
 key_len = 16 bytes
 data = "http://www.mamicode.com/Hi There"
 data_len = 8 bytes
 digest = 0x9294727a3638bb1c13f48ef8158bfc9d
  
 key = "Jefe"
 data = "http://www.mamicode.com/what do ya want for nothing?"
 data_len = 28 bytes
 digest = 0x750c783e6ab0b503eaa86e310a5db738
  
 key = 0xAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
  
  
  
 Krawczyk, et. al. Informational [Page 9]
 
 RFC 2104 HMAC February 1997
  
  
 key_len 16 bytes
 data = http://www.mamicode.com/0xDDDDDDDDDDDDDDDDDDDD...
 ..DDDDDDDDDDDDDDDDDDDD...
 ..DDDDDDDDDDDDDDDDDDDD...
 ..DDDDDDDDDDDDDDDDDDDD...
 ..DDDDDDDDDDDDDDDDDDDD
 data_len = 50 bytes
 digest = 0x56be34521d144c88dbb8c733f0e8b3f6
  
 Acknowledgments
  
 Pau-Chen Cheng, Jeff Kraemer, and Michael Oehler, have provided
 useful comments on early drafts, and ran the first interoperability
 tests of this specification. Jeff and Pau-Chen kindly provided the
 sample code and test vectors that appear in the appendix. Burt
 Kaliski, Bart Preneel, Matt Robshaw, Adi Shamir, and Paul van
 Oorschot have provided useful comments and suggestions during the
 investigation of the HMAC construction.
  
 References
  
 [ANSI] ANSI X9.9, "American National Standard for Financial
 Institution Message Authentication (Wholesale)," American
 Bankers Association, 1981. Revised 1986.
  
 [Atk] Atkinson, R., "IP Authentication Header", RFC 1826, August
 1995.
  
 [BCK1] M. Bellare, R. Canetti, and H. Krawczyk,
 "Keyed Hash Functions and Message Authentication",
 Proceedings of Crypto‘96, LNCS 1109, pp. 1-15.
 (http://www.research.ibm.com/security/keyed-md5.html)
  
 [BCK2] M. Bellare, R. Canetti, and H. Krawczyk,
 "Pseudorandom Functions Revisited: The Cascade Construction",
 Proceedings of FOCS‘96.
  
 [Dobb] H. Dobbertin, "The Status of MD5 After a Recent Attack",
 RSA Labs‘ CryptoBytes, Vol. 2 No. 2, Summer 1996.
 http://www.rsa.com/rsalabs/pubs/cryptobytes.html
  
 [PV] B. Preneel and P. van Oorschot, "Building fast MACs from hash
 functions", Advances in Cryptology -- CRYPTO‘95 Proceedings,
 Lecture Notes in Computer Science, Springer-Verlag Vol.963,
 1995, pp. 1-14.
  
 [MD5] Rivest, R., "The MD5 Message-Digest Algorithm",
 RFC 1321, April 1992.
  
  
  
 Krawczyk, et. al. Informational [Page 10]
 
 RFC 2104 HMAC February 1997
  
  
 [MM] Meyer, S. and Matyas, S.M., Cryptography, New York Wiley,
 1982.
  
 [RIPEMD] H. Dobbertin, A. Bosselaers, and B. Preneel, "RIPEMD-160: A
 strengthened version of RIPEMD", Fast Software Encryption,
 LNCS Vol 1039, pp. 71-82.
 ftp://ftp.esat.kuleuven.ac.be/pub/COSIC/bosselae/ripemd/.
  
 [SHA] NIST, FIPS PUB 180-1: Secure Hash Standard, April 1995.
  
 [Tsu] G. Tsudik, "Message authentication with one-way hash
 functions", In Proceedings of Infocom‘92, May 1992.
 (Also in "Access Control and Policy Enforcement in
 Internetworks", Ph.D. Dissertation, Computer Science
 Department, University of Southern California, April 1991.)
  
 [VW] P. van Oorschot and M. Wiener, "Parallel Collision
 Search with Applications to Hash Functions and Discrete
 Logarithms", Proceedings of the 2nd ACM Conf. Computer and
 Communications Security, Fairfax, VA, November 1994.
  
 Authors‘ Addresses
  
 Hugo Krawczyk
 IBM T.J. Watson Research Center
 P.O.Box 704
 Yorktown Heights, NY 10598
  
 EMail: hugo@watson.ibm.com
  
 Mihir Bellare
 Dept of Computer Science and Engineering
 Mail Code 0114
 University of California at San Diego
 9500 Gilman Drive
 La Jolla, CA 92093
  
 EMail: mihir@cs.ucsd.edu
  
 Ran Canetti
 IBM T.J. Watson Research Center
 P.O.Box 704
 Yorktown Heights, NY 10598
  
 EMail: canetti@watson.ibm.com
  
  
  
  
  
  
 Krawczyk, et. al. Informational [Page 11]
 
  

HMAC


联系
我们