This is a purely informative rendering of an RFC that includes verified errata. This rendering may not be used as a reference.

The following 'Verified' errata have been incorporated in this document: EID 4617
Network Working Group                                        D. McDonald
Request for Comments: 1751                                           NRL
Category: Informational                                    December 1994

              A Convention for Human-Readable 128-bit Keys

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
   the memo is unlimited.


   The Internet community has begun to address matters of security.
   Recent standards, including version 2 of SNMP [GM93], have explicit
   requirements for an authentication mechanism.  These require use of a
   keyed message-digest algorithm, MD5 [Riv92], with a key size of 128-
   bits.  A 128-bit key, while sufficiently strong, is hard for most
   people to read, remember, and type in.  This memo proposes a
   convention for use with Internet applications & protocols using 128-
   bit cryptographic keys.

A Solution Already Exists

   The S/Key(tm) one-time password system [Hal94] uses MD4 (and now MD5,
   as well) to compute one-time passwords.  It takes the 128-bit result
   of MD4 and collapses it to a 64-bit result.  Despite the size
   reduction, 64-bit one-time passwords are still difficult for ordinary
   people to remember and enter.  The authors of S/Key devised a system
   to make the 64-bit one-time password easy for people to enter.

   Their idea was to transform the password into a string of small
   English words.  English words are significantly easier for people to
   both remember and type.  The authors of S/Key started with a
   dictionary of 2048 English words, ranging in length from one to four
   characters.  The space covered by a 64-bit key (2^64) could be
   covered by six words from this dictionary (2^66) with room remaining
   for parity.  For example, an S/Key one-time password of hex value:

                            EB33 F77E E73D 4053

   would become the following six English words:

                       TIDE ITCH SLOW REIN RULE MOT

   Because of the need for interoperability, it is undesirable to have
   different dictionaries for different languages.  Also, the current
   dictionary only uses characters from the invariant portion of ISO-
   646.  Finally, there is an installed base of users and applications
   with this dictionary.

The Proposal

   The code (see Appendix A) which S/Key uses to convert 64-bit numbers
   to six English words contains two primitives which perform
   conversions either way.  The primitive btoe(char *engout,char *c)
   takes a 64-bit quantity referenced by c and places English words in
   the string referenced by engout.  The primitive etob(char *out,char
   *e) performs the opposite with an input string of English words
   referenced by e, and by placing the 64-bit result into the buffer
   referenced by out.

   The aforementioned primitives can be applied to both halves of a
   128-bit key, or both halves of a string of twelve English words.  Two
   new primitives (see Appendix B), key2eng(char *engout,char *key) and
   eng2key(char *keyout,char *eng) serve as wrappers which call the
   S/Key primitives twice, once for each half of the 128-bit key or
   string of twelve words.

   For example, the 128-bit key of:

                  CCAC 2AED 5910 56BE 4F90 FD44 1C53 4766

   would become


   Likewise, a user should be able to type in


   as a key, and the machine should make the translation to:

                  EFF8 1F9B FBC6 5350 920C DD74 16DE 8009

   If this proposal is to work, it is critical that the dictionary of
   English words does not change with different implementations.  A
   freely redistributable reference implementation is given in
   Appendices A and B.

Security Considerations

   This document recommends a method of representing 128-bit keys using
   strings of English words.  Since the strings of English words are
   easy to remember, people may potentially construct easy-to-guess
   strings of English words.  With easy-to-guess strings comes the
   possibility of a sentential equivalent of a dictionary attack.  In
   order to maximize the strength of any authentication mechanism that
   uses 128-bit keys, the keys must be sufficiently obscure.  In
   particular, people should avoid the temptation to devise sentences.


   S/Key is a registered trademark of Bell Communications Research.

   Thanks to Randall Atkinson for the bulk of the security
   considerations section, and for general advice.  Thanks to Phil Karn
   and Neil Haller for producing the S/Key one-time password system,
   which inspired this document.


   [GM93]  Galvin, J. and K. McCloghrie, "Security Protocols for version
   2 of the Simple Network Management Protocol (SNMPv2)", RFC 1446,
   Trusted Information Systems, Hughes LAN Systems, April 1993.

   [Hal94]  Haller, N., "The S/Key(tm) One-Time Password System",
   Proceedings of the Symposium on Network & Distributed Systems
   Security, Internet Society, San Diego, February 1994.

   [Riv92]  Rivest, R., "The MD5 Message-Digest Algorithm", RFC 1321,
   MIT Laboratory for Computer Science and RSA Data Security, Inc.,
   April 1992.

Author's Address

   Daniel L. McDonald
   United States Naval Research Laboratory
   Code 5544
   4555 Overlook Ave. SW
   Washington, DC 20375

   Phone:  (202) 404-7122

Appendix A - Source for S/Key 8-bytes to/from Words Routines (put.c)

/* This code originally appeared in the source for S/Key(TM),
 * available in the directory
 * It has been modified only to remove explicit S/Key(TM) references.

#include <stdio.h>
#include <string.h>
#include <assert.h>
#include <ctype.h>

#ifdef __STDC__
#define __ARGS(x) x
#define __ARGS(x) ()

static unsigned long extract __ARGS((char *s,int start,int length));
static void standard __ARGS((char *word));
static void insert __ARGS((char *s, int x, int start, int length));
static int wsrch __ARGS((char *w,int low,int high));

/* Dictionary for integer-word translations */
char Wp[2048][4] = { "A", "ABE", "ACE", "ACT", "AD", "ADA", "ADD",
"AGO", "AID", "AIM", "AIR", "ALL", "ALP", "AM", "AMY", "AN", "ANA",
"AND", "ANN", "ANT", "ANY", "APE", "APS", "APT", "ARC", "ARE", "ARK",
"ARM", "ART", "AS", "ASH", "ASK", "AT", "ATE", "AUG", "AUK", "AVE",
"AWE", "AWK", "AWL", "AWN", "AX", "AYE", "BAD", "BAG", "BAH", "BAM",
"BAN", "BAR", "BAT", "BAY", "BE", "BED", "BEE", "BEG", "BEN", "BET",
"BEY", "BIB", "BID", "BIG", "BIN", "BIT", "BOB", "BOG", "BON", "BOO",
"BOP", "BOW", "BOY", "BUB", "BUD", "BUG", "BUM", "BUN", "BUS", "BUT",
"BUY", "BY", "BYE", "CAB", "CAL", "CAM", "CAN", "CAP", "CAR", "CAT",
"CAW", "COD", "COG", "COL", "CON", "COO", "COP", "COT", "COW", "COY",
"CRY", "CUB", "CUE", "CUP", "CUR", "CUT", "DAB", "DAD", "DAM", "DAN",
"DAR", "DAY", "DEE", "DEL", "DEN", "DES", "DEW", "DID", "DIE", "DIG",
"DIN", "DIP", "DO", "DOE", "DOG", "DON", "DOT", "DOW", "DRY", "DUB",
"DUD", "DUE", "DUG", "DUN", "EAR", "EAT", "ED", "EEL", "EGG", "EGO",
"ELI", "ELK", "ELM", "ELY", "EM", "END", "EST", "ETC", "EVA", "EVE",
"EWE", "EYE", "FAD", "FAN", "FAR", "FAT", "FAY", "FED", "FEE", "FEW",
"FIB", "FIG", "FIN", "FIR", "FIT", "FLO", "FLY", "FOE", "FOG", "FOR",
"FRY", "FUM", "FUN", "FUR", "GAB", "GAD", "GAG", "GAL", "GAM", "GAP",
"GAS", "GAY", "GEE", "GEL", "GEM", "GET", "GIG", "GIL", "GIN", "GO",
"GOT", "GUM", "GUN", "GUS", "GUT", "GUY", "GYM", "GYP", "HA", "HAD",
"HAL", "HAM", "HAN", "HAP", "HAS", "HAT", "HAW", "HAY", "HE", "HEM",
"HEN", "HER", "HEW", "HEY", "HI", "HID", "HIM", "HIP", "HIS", "HIT",
"HO", "HOB", "HOC", "HOE", "HOG", "HOP", "HOT", "HOW", "HUB", "HUE",
"HUG", "HUH", "HUM", "HUT", "I", "ICY", "IDA", "IF", "IKE", "ILL",
"INK", "INN", "IO", "ION", "IQ", "IRA", "IRE", "IRK", "IS", "IT", "ITS",
"IVY", "JAB", "JAG", "JAM", "JAN", "JAR", "JAW", "JAY", "JET", "JIG",
"JIM", "JO", "JOB", "JOE", "JOG", "JOT", "JOY", "JUG", "JUT", "KAY",
"KEG", "KEN", "KEY", "KID", "KIM", "KIN", "KIT", "LA", "LAB", "LAC",
"LAD", "LAG", "LAM", "LAP", "LAW", "LAY", "LEA", "LED", "LEE", "LEG",
"LEN", "LEO", "LET", "LEW", "LID", "LIE", "LIN", "LIP", "LIT", "LO",
"LOB", "LOG", "LOP", "LOS", "LOT", "LOU", "LOW", "LOY", "LUG", "LYE",
"MA", "MAC", "MAD", "MAE", "MAN", "MAO", "MAP", "MAT", "MAW", "MAY",
"ME", "MEG", "MEL", "MEN", "MET", "MEW", "MID", "MIN", "MIT", "MOB",
"MOD", "MOE", "MOO", "MOP", "MOS", "MOT", "MOW", "MUD", "MUG", "MUM",
"MY", "NAB", "NAG", "NAN", "NAP", "NAT", "NAY", "NE", "NED", "NEE",
"NET", "NEW", "NIB", "NIL", "NIP", "NIT", "NO", "NOB", "NOD", "NON",
"NOR", "NOT", "NOV", "NOW", "NU", "NUN", "NUT", "O", "OAF", "OAK",
"OAR", "OAT", "ODD", "ODE", "OF", "OFF", "OFT", "OH", "OIL", "OK",
"OLD", "ON", "ONE", "OR", "ORB", "ORE", "ORR", "OS", "OTT", "OUR",
"OUT", "OVA", "OW", "OWE", "OWL", "OWN", "OX", "PA", "PAD", "PAL",
"PAM", "PAN", "PAP", "PAR", "PAT", "PAW", "PAY", "PEA", "PEG", "PEN",
"PEP", "PER", "PET", "PEW", "PHI", "PI", "PIE", "PIN", "PIT", "PLY",
"PO", "POD", "POE", "POP", "POT", "POW", "PRO", "PRY", "PUB", "PUG",
"PUN", "PUP", "PUT", "QUO", "RAG", "RAM", "RAN", "RAP", "RAT", "RAW",
"RAY", "REB", "RED", "REP", "RET", "RIB", "RID", "RIG", "RIM", "RIO",
"RIP", "ROB", "ROD", "ROE", "RON", "ROT", "ROW", "ROY", "RUB", "RUE",
"RUG", "RUM", "RUN", "RYE", "SAC", "SAD", "SAG", "SAL", "SAM", "SAN",
"SAP", "SAT", "SAW", "SAY", "SEA", "SEC", "SEE", "SEN", "SET", "SEW",
"SHE", "SHY", "SIN", "SIP", "SIR", "SIS", "SIT", "SKI", "SKY", "SLY",
"SO", "SOB", "SOD", "SON", "SOP", "SOW", "SOY", "SPA", "SPY", "SUB",
"SUD", "SUE", "SUM", "SUN", "SUP", "TAB", "TAD", "TAG", "TAN", "TAP",
"TAR", "TEA", "TED", "TEE", "TEN", "THE", "THY", "TIC", "TIE", "TIM",
"TIN", "TIP", "TO", "TOE", "TOG", "TOM", "TON", "TOO", "TOP", "TOW",
"TOY", "TRY", "TUB", "TUG", "TUM", "TUN", "TWO", "UN", "UP", "US",
"USE", "VAN", "VAT", "VET", "VIE", "WAD", "WAG", "WAR", "WAS", "WAY",
"WE", "WEB", "WED", "WEE", "WET", "WHO", "WHY", "WIN", "WIT", "WOK",
"WON", "WOO", "WOW", "WRY", "WU", "YAM", "YAP", "YAW", "YE", "YEA",
"YES", "YET", "YOU", "ABED", "ABEL", "ABET", "ABLE", "ABUT", "ACHE",

/* Encode 8 bytes in 'c' as a string of English words.
 * Returns a pointer to a static buffer
char *
char *c, *engout;
        char cp[10];     /* add in room for the parity 2 bits*/
EID 4617 (Verified) is as follows:

Section: Appendix A

Original Text:

char *c, *engout;
        char cp[9];     /* add in room for the parity 2 bits*/

Corrected Text:

char *c, *engout;
        char cp[10];     /* add in room for the parity 2 bits*/
This is an off-by-one error in the sample code in Appendix A.

Further down, we have this loop:
for(p = 0,i = 0; i < 64;i += 2)
p += extract(cp,i,2);

So i goes all the way to 62, and 9-byte cp is passed to extract()

In extract, we have this:
static unsigned long
extract(s, start, length)
char *s;
int start, length;
cr = s[start/8 +2];

If start is 62, then (start/8+2) is 9. s is the same 9-byte cp, and s[start/8 +2] is a one-byte overflow.
int p,i ; engout[0] = '\0'; memcpy(cp, c,8); /* compute parity */ for(p = 0,i = 0; i < 64;i += 2) p += extract(cp,i,2); cp[8] = (char)p << 6; strncat(engout,&Wp[extract(cp, 0,11)][0],4); strcat(engout," "); strncat(engout,&Wp[extract(cp,11,11)][0],4); strcat(engout," "); strncat(engout,&Wp[extract(cp,22,11)][0],4); strcat(engout," "); strncat(engout,&Wp[extract(cp,33,11)][0],4); strcat(engout," "); strncat(engout,&Wp[extract(cp,44,11)][0],4); strcat(engout," "); strncat(engout,&Wp[extract(cp,55,11)][0],4); #ifdef notdef printf("engout is %s\n\r",engout); #endif return(engout); } /* convert English to binary * returns 1 OK - all good words and parity is OK * 0 word not in data base * -1 badly formed in put ie > 4 char word * -2 words OK but parity is wrong */ int etob(out, e) char *out; char *e; { char *word; int i, p, v,l, low,high; char b[9]; char input[36]; if(e == NULL) return -1; strncpy(input,e,sizeof(input)); memset(b, 0, sizeof(b)); memset(out, 0, 8); for(i=0,p=0;i<6;i++,p+=11){ if((word = strtok(i == 0 ? input : NULL," ")) == NULL) return -1; l = strlen(word); if(l > 4 || l < 1){ return -1; } else if(l < 4){ low = 0; high = 570; } else { low = 571; high = 2047; } standard(word); if( (v = wsrch(word,low,high)) < 0 ) return 0; insert(b,v,p,11); } /* now check the parity of what we got */ for(p = 0, i = 0; i < 64; i +=2) p += extract(b, i, 2); if( (p & 3) != extract(b, 64,2) ) return -2; memcpy(out,b,8); return 1; } /* Display 8 bytes as a series of 16-bit hex digits */ char * put8(out,s) char *out; char *s; { sprintf(out,"%02X%02X %02X%02X %02X%02X %02X%02X", s[0] & 0xff,s[1] & 0xff,s[2] & 0xff, s[3] & 0xff,s[4] & 0xff,s[5] & 0xff, s[6] & 0xff,s[7] & 0xff); return out; } #ifdef notdef /* Encode 8 bytes in 'cp' as stream of ascii letters. * Provided as a possible alternative to btoe() */ char * btoc(cp) char *cp; { int i; static char out[31]; /* code out put by characters 6 bits each added to 0x21 (!)*/ for(i=0;i <= 10;i++){ /* last one is only 4 bits not 6*/ out[i] = '!'+ extract(cp,6*i,i >= 10 ? 4:6); } out[i] = '\0'; return(out); } #endif /* Internal subroutines for word encoding/decoding */ /* Dictionary binary search */ static int wsrch(w,low,high) char *w; int low, high; { int i,j; for(;;){ i = (low + high)/2; if((j = strncmp(w,Wp[i],4)) == 0) return i; /* Found it */ if(high == low+1){ /* Avoid effects of integer truncation in /2 */ if(strncmp(w,Wp[high],4) == 0) return high; else return -1; } if(low >= high) return -1; /* I don't *think* this can happen...*/ if(j < 0) high = i; /* Search lower half */ else low = i; /* Search upper half */ } } static void insert(s, x, start, length) char *s; int x; int start, length; { unsigned char cl; unsigned char cc; unsigned char cr; unsigned long y; int shift; assert(length <= 11); assert(start >= 0); assert(length >= 0); assert(start +length <= 66); shift = ((8 -(( start + length) % 8))%8); y = (long) x << shift; cl = (y >> 16) & 0xff; cc = (y >> 8) & 0xff; cr = y & 0xff; if(shift + length > 16){ s[start /8] |= cl; s[start/8 +1] |= cc; s[start/8 +2] |= cr; } else if(shift +length > 8){ s[start/8] |= cc; s[start/8 + 1] |= cr; } else { s[start/8] |= cr; } } static void standard(word) register char *word; { while(*word){ if(!isascii(*word)) break; if(islower(*word)) *word = toupper(*word); if(*word == '1') *word = 'L'; if(*word == '0') *word = 'O'; if(*word == '5') *word = 'S'; word++; } } /* Extract 'length' bits from the char array 's' starting with bit 'start' */ static unsigned long extract(s, start, length) char *s; int start, length; { unsigned char cl; unsigned char cc; unsigned char cr; unsigned long x; assert(length <= 11); assert(start >= 0); assert(length >= 0); assert(start +length <= 66); cl = s[start/8]; cc = s[start/8 +1]; cr = s[start/8 +2]; x = ((long)(cl<<8 | cc) <<8 | cr) ; x = x >> (24 - (length + (start %8))); x =( x & (0xffff >> (16-length) ) ); return(x); } Appendix B - Source for 128-bit key to/from English words (convert.c) /* convert.c -- Wrapper to S/Key binary-to-English routines. Daniel L. McDonald -- U. S. Naval Research Laboratory. */ #include <string.h> /* eng2key() assumes words must be separated by spaces only. eng2key() returns 1 if succeeded 0 if word not in dictionary -1 if badly formed string -2 if words are okay but parity is wrong. (see etob() in S/Key) */ int eng2key(keyout,eng) char *keyout,*eng; { int rc=0,state=1; char *eng2; /* Find pointer to word 7. */ for (eng2 = eng; rc<7 && (*(++eng2) != '\0'); ) if (*eng2 != ' ') { rc += state; state = 0; } else state=1; if ( (rc = etob(keyout,eng)) != 1) return rc; rc = etob(keyout+8,eng2); return rc; } /* key2eng() assumes string referenced by engout has at least 60 characters (4*12 + 11 spaces + '\0') of space. key2eng() returns pointer to engout. */ char *key2eng(engout,key) char *engout,*key; { btoe(engout,key); strcat(engout," "); btoe(engout+strlen(engout),key+8); return engout; }