1 { Copyright (C) 2005 Bas Steendijk and Peter Green
\r
2 For conditions of distribution and use, see copyright notice in zlib_license.txt
\r
3 which is included in the package
\r
4 ----------------------------------------------------------------------------- }
\r
10 {$include lcoreconfig.inc}
\r
13 written by Bas Steendijk (beware)
\r
15 the aim of this unit is to provide randomness in a consistent way, using OS specific methods for seeding
\r
17 this unit uses MD5 for performance and code size, but it is made so it is easy to use a different hash,
\r
18 as long as it is atleat 128 bits, and a multiple of the "word size" (32 bits)
\r
22 - for the code to be:
\r
23 - relatively simple and small
\r
26 - for the numbers to be
\r
27 - random: pass diehard and similar tests
\r
28 - unique: generate UUID's
\r
29 - secure: difficult for a remote attacker to guess the internal state, even
\r
30 when given some output
\r
32 typical intended uses:
\r
33 - anything that needs random numbers without extreme demands on security or
\r
34 speed should be able to use this
\r
35 - seeding other (faster) RNG's
\r
36 - generation of passwords, UUID's, cookies, and session keys
\r
37 - randomizing protocol fields to protect against spoofing attacks
\r
38 - randomness for games
\r
40 this is not intended to be directly used for:
\r
41 - high securirity purposes (generating RSA root keys etc)
\r
42 - needing random numbers at very high rates (disk wiping, some simulations, etc)
\r
45 - 24 MB/s on 2.2 GHz athlon64 core on windows XP 32 bits
\r
46 - 6.4 MB/s on 1 GHz p3 on linux
\r
49 - fpc 2.2, linux: fastmd5: 12 kb; lcorernd: 6 kb.
\r
50 - delphi 6: fastmd5: 3 kb; lcorernd: 2 kb
\r
52 reasoning behind the security of this RNG:
\r
55 1: i assume that any attacker has no local access to the machine. if one gained
\r
56 this, then there are more seriousness weaknesses to consider.
\r
57 2: i attempt to use enough seeding to be difficult to guess.
\r
58 on windows: GUID, various readouts of hi res timestamps, heap stats, cursor
\r
60 on *nix: i assume /dev/(u)random output is secure and difficult to guess. if
\r
61 it misses, i use /dev/wtmp, which typically has as lot of entropy in it. i also use hi res timestamps.
\r
62 3: on a state compromise, one can easily get up to the hash size worth of previous output, beyond that one has
\r
63 to invert the hash operation.
\r
65 - mixing/expansion: a secure hash random walk on a buffer with a constant secret and a changing exposed part,
\r
66 the big secret part serves to make it difficult for an attacker to predict next and previous output.
\r
67 the secret part is changed during a reseed.
\r
72 <wwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwww>
\r
73 ____________________________ ________________________________________________
\r
75 [hashsize][hashsize][hashsize]
\r
76 <rrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr>
\r
79 <wwwwwwwwwwwwwwwwww>
\r
80 <rrrrrrrrrrrrrrrrrrrrrrrrrrrr>
\r
84 [ output ][ secret ]
\r
87 this needs testing on platforms other than i386
\r
90 these routines are called by everything else in lcore, and if the app coder desires, by the app.
\r
91 because one may want to use their own random number source, the PRNG here can be excluded from linking,
\r
92 and the routines here can be hooked.
\r
95 {$include uint32.inc}
\r
97 {return a dword with 32 random bits}
\r
102 randomdword:function:wordtype;
\r
104 {fill a buffer with random bytes}
\r
105 procedure fillrandom(var buf;length:integer);
\r
107 {generate an integer of 0 <= N < i}
\r
108 function randominteger(i:longint):longint;
\r
110 {generate an integer with the lowest b bits being random}
\r
111 function randombits(b:integer):longint;
\r
113 {generate a version 4 random uuid}
\r
114 function generate_uuid:string;
\r
116 {$ifndef nolcorernd}
\r
118 {call this to mix seeding into the pool. is normally done automatically and does not have to be called
\r
119 but can be done if one desires more security, for example for key generation}
\r
120 procedure seedpool;
\r
122 {get some raw OS specific randomness. the output is not mixed so it should not be used directly as random numbers}
\r
123 function collect_seeding(var output;const bufsize:integer):integer;
\r
125 function internalrandomdword:wordtype;
\r
128 reseedinterval:integer=64;
\r
133 {$ifndef nolcorernd}
\r
135 {$ifdef win32}windows,activex,{$endif}
\r
140 baseunix,unix,unixutil,
\r
145 {$ifdef unix}{$include unixstuff.inc}{$endif}
\r
148 {hashtype must be array of bytes}
\r
153 wordsize=1 shl wordsizeshift;
\r
154 //wordsize check commented out for d3 compatibility
\r
155 //{ $if (wordsize <> sizeof(wordtype))}'wordsizeshift must be setcorrectly'{ $ifend}
\r
156 hashsize=sizeof(hashtype);
\r
157 halfhashsize=hashsize div 2;
\r
158 hashdwords=hashsize div wordsize;
\r
159 pooldwords=3*hashdwords;
\r
161 hashpasssize=48; {this number has to be small enough that hashing this size uses only one block transform}
\r
164 {the seed part of this buffer must be atleast as big as the OS seed (windows: 104 bytes, unix: 36 bytes)}
\r
165 pool:array[0..(pooldwords+seeddwords-1)] of wordtype;
\r
166 reseedcountdown:integer;
\r
169 function collect_seeding(var output;const bufsize:integer):integer;
\r
172 guid:array[0..3] of longint;
\r
173 qpcbuf:array[0..1] of longint;
\r
174 rdtscbuf:array[0..1] of longint;
\r
175 systemtimebuf:array[0..3] of longint;
\r
180 end absolute output;
\r
181 rdtsc_0,rdtsc_1:integer;
\r
184 if (bufsize < sizeof(l)) then exit;
\r
185 result := sizeof(l);
\r
187 l.pid := GetCurrentProcessId;
\r
188 l.tid := GetCurrentThreadId;
\r
191 cocreateguid(tguid(l.guid));
\r
193 {QUERYPERFORMANCECOUNTER}
\r
194 queryperformancecounter(tlargeinteger(l.qpcbuf));
\r
203 l.rdtscbuf[0] := rdtsc_0;
\r
204 l.rdtscbuf[1] := rdtsc_1;
\r
207 getsystemtime(tsystemtime(l.systemtimebuf));
\r
210 getcursorpos(l.cursor);
\r
212 l.hs := getheapstatus;
\r
219 wtmpinited:boolean;
\r
220 wtmpcached:hashtype;
\r
222 procedure wtmphash;
\r
225 buf:array[0..4095] of byte;
\r
229 if wtmpinited then exit;
\r
231 assignfile(f,'/var/log/wtmp');
\r
233 {$i-}reset(f,1);{$i+}
\r
234 if (ioresult <> 0) then exit;
\r
236 while not eof(f) do begin
\r
237 blockread(f,buf,sizeof(buf),numread);
\r
238 md5process(state,buf,numread);
\r
241 md5finish(state,wtmpcached);
\r
242 wtmpinited := true;
\r
246 function collect_seeding(var output;const bufsize:integer):integer;
\r
251 devrnd:array[0..3] of integer;
\r
252 rdtscbuf:array[0..1] of integer;
\r
255 end absolute output;
\r
256 rdtsc_0,rdtsc_1:integer;
\r
260 if (bufsize < sizeof(l)) then exit;
\r
261 result := sizeof(l);
\r
265 assignfile(f,'/dev/urandom');
\r
267 {$i-}reset(f,1);{$i+}
\r
269 if (a <> 0) then begin
\r
270 assignfile(f,'/dev/random');
\r
271 {$i-}reset(f,1);{$i+}
\r
274 if (a = 0) then begin
\r
275 blockread(f,l.devrnd,sizeof(l.devrnd));
\r
278 {the OS we are on has no /dev/random or /dev/urandom, get a hash from /var/log/wtmp}
\r
280 move(wtmpcached,l.devrnd,sizeof(l.devrnd));
\r
282 {get more randomness in case there's no /dev/random}
\r
283 {$ifdef cpu386}{$ASMMODE intel}
\r
289 l.rdtscbuf[0] := rdtsc_0;
\r
290 l.rdtscbuf[1] := rdtsc_1;
\r
293 gettimeofday(l.tv);
\r
298 {this produces a hash which is twice the native hash size (32 bytes for MD5)}
\r
299 procedure bighash(const input;len:integer;var output);
\r
301 inarr:array[0..65535] of byte absolute input;
\r
302 outarr:array[0..65535] of byte absolute output;
\r
304 h1,h2,h3,h4:hashtype;
\r
309 getmd5(inarr[0],a,h1);
\r
310 getmd5(inarr[a],len-a,h2);
\r
312 move(h1[0],h3[0],halfhashsize);
\r
313 move(h2[0],h3[halfhashsize],halfhashsize);
\r
314 move(h1[halfhashsize],h4[0],halfhashsize);
\r
315 move(h2[halfhashsize],h4[halfhashsize],halfhashsize);
\r
317 getmd5(h3,hashsize,outarr[0]);
\r
318 getmd5(h4,hashsize,outarr[hashsize]);
\r
321 procedure seedpool;
\r
325 a := collect_seeding(pool[pooldwords],seeddwords*wordsize);
\r
326 if (a = 0) then halt;
\r
327 bighash(pool[hashdwords],(2*hashsize)+a,pool[hashdwords]);
\r
328 getmd5(pool[0],hashpasssize,pool[0]);
\r
331 function internalrandomdword;
\r
333 if (reseedcountdown <= 0) then begin
\r
335 reseedcountdown := reseedinterval * hashdwords;
\r
336 end else if ((reseedcountdown mod hashdwords) = 0) then begin;
\r
337 getmd5(pool[0],hashpasssize,pool[0]);
\r
339 dec(reseedcountdown);
\r
341 result := pool[reseedcountdown mod hashdwords];
\r
345 procedure fillrandom(var buf;length:integer);
\r
348 buf_:array[0..16383] of uint32 absolute buf;
\r
352 for a := (length shr wordsizeshift)-1 downto 0 do begin
\r
353 buf_[b] := randomdword;
\r
356 length := length and (wordsize-1);
\r
357 if length <> 0 then begin
\r
359 move(a,buf_[b],length);
\r
366 function randombits(b:integer):longint;
\r
368 result := randomdword;
\r
369 result := result and (-1 shr (wordsizebits-b));
\r
370 if (b = 0) then result := 0;
\r
373 function randominteger(i:longint):longint;
\r
378 //bitscounter := bitscounter + numofbitsininteger(i);
\r
379 if (i = 0) then begin
\r
383 {find number of bits needed}
\r
385 if (j < 0) then begin
\r
386 result := randombits(wordsizebits);
\r
388 end else if (j >= (1 shl (wordsizebits-2))) then begin
\r
389 b := wordsizebits-1
\r
392 for a := 0 to (wordsizebits-2) do begin
\r
393 if j < 1 shl a then begin
\r
400 result := randombits(b);
\r
405 ch:array[0..15] of char='0123456789abcdef';
\r
407 function generate_uuid:string;
\r
409 buf:array[0..7] of word;
\r
410 function inttohex(w:word):string;
\r
412 result := ch[w shr 12] + ch[(w shr 8) and $f] + ch[(w shr 4) and $f] + ch[w and $f];
\r
415 fillrandom(buf,sizeof(buf));
\r
418 buf[3] := (buf[3] and $fff) or $4000;
\r
421 buf[4] := (buf[4] and $3fff) or $8000;
\r
423 result := inttohex(buf[0]) + inttohex(buf[1]) + '-' + inttohex(buf[2]) +'-'+ inttohex(buf[3]) + '-' + inttohex(buf[4])
\r
424 + '-' + inttohex(buf[5]) + inttohex(buf[6]) + inttohex(buf[7]);
\r
427 {$ifndef nolcorernd}
\r
428 initialization randomdword := @internalrandomdword;
\r