|
|
|
1 |
#!/usr/bin/env python |
|
2 |
# -*- coding: iso-8859-1 -*- |
|
3 |
|
|
4 |
from __future__ import division |
|
5 |
from math import * |
|
6 |
import Crypto.Util.number as cn |
|
7 |
|
|
8 |
def log2(x): |
|
9 |
return log(x, 2) |
|
10 |
|
|
11 |
def logint(x, base, __cache=[(8, 2), 3]): |
|
12 |
"Return int(ceil(log(x, base))) without rounding errors." |
|
13 |
# Rounding error example: |
|
14 |
#>>> log(2**96, 2**12) |
|
15 |
#8.0000000000000018 |
|
16 |
if (x, base) == __cache[0]: |
|
17 |
return __cache[1] |
|
18 |
c = int(ceil(log(x, base))) # This works most of the time. |
|
19 |
while x > base**c: |
|
20 |
c += 1 # If not, this will fix it. |
|
21 |
while x <= base**(c-1): |
|
22 |
c -= 1 # Or this. |
|
23 |
__cache[:] = (x, base), c |
|
24 |
return c |
|
25 |
|
|
26 |
def istwopower(x, __cache={}): |
|
27 |
c = __cache.get(x) |
|
28 |
if c is not None: |
|
29 |
return c |
|
30 |
c = 0L # This must be a long, or 1<<c will lose bits. |
|
31 |
while x > 1<<c: |
|
32 |
c += 1 |
|
33 |
if x != 1<<c: |
|
34 |
return None |
|
35 |
__cache[x] = c |
|
36 |
return c |
|
37 |
|
|
38 |
def int2list(x, xmax=None, s=256): |
|
39 |
"""Return the integer x as a little-endian list of bytes with s |
|
40 |
states each. s is the number of states, not the number of bits. |
|
41 |
If xmax is given, the returned list will be large enough to |
|
42 |
contain xmax-1.""" |
|
43 |
if xmax is None: |
|
44 |
xmax = x+1 |
|
45 |
assert 0 <= x < xmax |
|
46 |
l = [0] * logint(xmax, s) |
|
47 |
c = istwopower(s) |
|
48 |
if c is None: |
|
49 |
for i in xrange(len(l)): |
|
50 |
l[i] = int(x%s) |
|
51 |
x = x//s |
|
52 |
else: # s is exactly 2**c==1<<c so we can optimize somewhat. |
|
53 |
mask = (1<<c)-1 |
|
54 |
for i in xrange(len(l)): |
|
55 |
l[i] = int(x&mask) |
|
56 |
x = x>>c |
|
57 |
return l |
|
58 |
|
|
59 |
def list2int(l, s=256): |
|
60 |
"""Return list interpreted as a little-endian list of bytes with s |
|
61 |
states each. s is the number of states, not the number of bits.""" |
|
62 |
while len(l) > 1: |
|
63 |
l[-2] += l[-1]*s |
|
64 |
del l[-1] |
|
65 |
return l[0] |
|
66 |
|
|
67 |
def nextprime(begin=1, __cache=[7,7]): |
|
68 |
"""Return the lowest prime >= begin.""" |
|
69 |
if begin == __cache[0]: |
|
70 |
return __cache[1] |
|
71 |
n = begin |
|
72 |
if not n%2: n+=1 |
|
73 |
while not cn.isPrime(long(n)): |
|
74 |
n += 2 |
|
75 |
__cache[:] = begin, n |
|
76 |
return int(n) |
|
77 |
|
|
78 |
# The following functions were first described by J. Lawrence Carter |
|
79 |
# and Mark N. Wegman in their papers "Universal classes of hash |
|
80 |
# functions" (1979) and "New hash functions and their use in |
|
81 |
# authentication and set equality" (1981). |
|
82 |
# All functions try to follow the Wegman-Carter papers as closely as |
|
83 |
# possible and all quotes are from the papers. |
|
84 |
|
|
85 |
def H1(a, b, key=None): |
|
86 |
"""Return a member of the hash family H_1 from Wegman-Carter 1979 |
|
87 |
Inputs: |
|
88 |
a -- Number of input states |
|
89 |
b -- Number of output states |
|
90 |
key -- a tuple (p, m, n): |
|
91 |
p -- A (non-secret) prime larger than or equal to a |
|
92 |
q -- Half of the secret key. 0< q |
|
93 |
r -- Half of the secret key. 0<=r |
|
94 |
Output: |
|
95 |
Normal mode: |
|
96 |
A hash function mapping range(a) to range(b) |
|
97 |
Parameter mode: |
|
98 |
If H1 is called with no key, the smallest possible p is |
|
99 |
returned to ease construction of a key. |
|
100 |
""" |
|
101 |
if key is None: |
|
102 |
return nextprime(a) |
|
103 |
p, q, r = key |
|
104 |
assert (a>b) and (p>=a) and (0and (0<=r |
|
105 |
def g(x): # "A natural choice for g is the residue modulo b." |
|
106 |
return x % b |
|
107 |
def h(m): |
|
108 |
return (q*m+r) % p |
|
109 |
def f(m): |
|
110 |
# Docstring is set below |
|
111 |
assert 0<=m |
|
112 |
return g(h(m)) |
|
113 |
f.__doc__ = \ |
|
114 |
"""This is a hash function from the hash family H_1 from |
|
115 |
Wegman-Carter 1979 using the key %s. |
|
116 |
Inputs: |
|
117 |
m -- The integer to be hashed in range(%d) |
|
118 |
Output: |
|
119 |
A hash value in range(%d) |
|
120 |
""" % (str(key), a, b) |
|
121 |
return f |
|
122 |
|
|
123 |
def Hprime(aprime, bprime, flist=None): |
|
124 |
"""Return a member of the hash family H' from Wegman-Carter 1980 |
|
125 |
Inputs: |
|
126 |
aprime -- Number of input bits |
|
127 |
bprime -- Number of output states |
|
128 |
flist -- A sequence of secret hash functions |
|
129 |
Output: |
|
130 |
Normal mode: |
|
131 |
A hash function mapping range(2**aprime) to range(2**bprime) |
|
132 |
Parameter mode: |
|
133 |
Hprime needs a sequence flist of hash functions from a |
|
134 |
universal_2 family. The number of functions, input states and |
|
135 |
output states are dependent on the inner workings of Hprime |
|
136 |
and need not be exposed to the outside. Instead, if Hprime is |
|
137 |
called without flist those specifications are returned as a |
|
138 |
tuple (a, b, len_f): |
|
139 |
a -- Number of input states of each hash function |
|
140 |
b -- Number of output states of each hash function |
|
141 |
len_f -- Number of hash functions in flist |
|
142 |
""" |
|
143 |
s = bprime + int(ceil(log2(log2(aprime)))) |
|
144 |
# "Let H be some strongly universal_2 class of functions which map |
|
145 |
# bit strings of length 2s to ones of length s" |
|
146 |
a = 2**(2*s) |
|
147 |
b = 2**( s) |
|
148 |
# The "or 1" is needed because the length calculation in W-C-80 |
|
149 |
# doesn't account for the extra padding when the message is |
|
150 |
# smaller than s from the beginning. |
|
151 |
len_f = int(ceil(log2(ceil(aprime/s)))) or 1 |
|
152 |
if flist is None: |
|
153 |
return (a, b, len_f) |
|
154 |
assert len(flist) == len_f |
|
155 |
def f(substrings, hashfunction): |
|
156 |
# Apply hashfunction to all substrings and concatenate pairwise |
|
157 |
for i in xrange(len(substrings)//2): |
|
158 |
substrings[i:i+2] = [hashfunction(substrings[i ]) + |
|
159 |
hashfunction(substrings[i+1]) * 2**s] |
|
160 |
if len(substrings)%2: |
|
161 |
substrings[-1] = hashfunction(substrings[-1]) |
|
162 |
def fprime(m): |
|
163 |
# Docstring is set below |
|
164 |
assert 0 <= m < 2**aprime |
|
165 |
# "The message is broken into substrings of length 2s." |
|
166 |
substrings = int2list(m, xmax=2**aprime, s=2**(2*s)) |
|
167 |
for f_i in flist[:-1]: |
|
168 |
assert len(substrings) > 1 |
|
169 |
f(substrings, f_i) |
|
170 |
# "This process is repeated using f_2,f_3,... until only one |
|
171 |
# substring of length s is left." |
|
172 |
assert len(substrings) == 1 |
|
173 |
substring = flist[-1](substrings[0]) |
|
174 |
assert 0 <= substring < 2**s |
|
175 |
# "The tag is the low-order b' bits of this substring." |
|
176 |
return substring % 2**bprime |
|
177 |
fprime.__doc__ = "This is a hash function from the hash family H' " \ |
|
178 |
"from Wegman-Carter 1980.\n" \ |
|
179 |
" Inputs:\n" \ |
|
180 |
" m -- A %d bit integer to be hashed\n" \ |
|
181 |
" Output:\n" \ |
|
182 |
" A %d bit hash value\n" \ |
|
183 |
% (aprime, bprime) |
|
184 |
return fprime |
|
185 |
|
|
186 |
def Hprime_H1(aprime, bprime, key=None): |
|
187 |
"""Return a member of the hash family H' from Wegman-Carter 1980 |
|
188 |
using the sub-hash family H_1 from Wegman-Carter 1979 |
|
189 |
Inputs: |
|
190 |
aprime -- Number of input bits |
|
191 |
bprime -- Number of output bits |
|
192 |
key -- A secret key |
|
193 |
Outputs: |
|
194 |
Normal mode: |
|
195 |
A hash function mapping range(2**aprime) to range(2**bprime) |
|
196 |
Parameter mode: |
|
197 |
If Hprime_H1 is called without a key an integer maxkey is |
|
198 |
returned. The key should be in range(maxkey). |
|
199 |
""" |
|
200 |
# Get key parameters |
|
201 |
a, b, len_f = Hprime(aprime, bprime) |
|
202 |
p = H1(a, b) |
|
203 |
maxkey = ((p-1)*p)**len_f |
|
204 |
if key is None: |
|
205 |
return maxkey |
|
206 |
assert 0 <= key < maxkey |
|
207 |
flist = [] |
|
208 |
for thiskey in int2list(key, xmax=maxkey, s=(p-1)*p): |
|
209 |
q, r = divmod(thiskey, p) |
|
210 |
q += 1 |
|
211 |
flist.append(H1(a, b, (p, q, r))) |
|
212 |
return Hprime(aprime, bprime, flist) |
|
213 |
|
|
214 |
def Hprime_H1_compact(aprime, bprime, key=None): |
|
215 |
"""A more compact implementation of Wegman-Carter 1980 with H1 |
|
216 |
from W-C 1979. |
|
217 |
The functions above are written to mimic the language of |
|
218 |
Wegman-Carter as much as possible. Sometimes it might be easier to |
|
219 |
understand a more compact language. This code should do exactly |
|
220 |
the same as the one above, but in far less lines and with no error |
|
221 |
checking. It is approximately three times faster than Hprime_H1(). |
|
222 |
Inputs: |
|
223 |
aprime -- Number of input bits |
|
224 |
bprime -- Number of output bits |
|
225 |
key -- A secret key |
|
226 |
Outputs: |
|
227 |
Normal mode: |
|
228 |
A hash function mapping range(2**aprime) to range(2**bprime) |
|
229 |
Parameter mode: |
|
230 |
If Hprime_H1_compact is called without a key an integer maxkey |
|
231 |
is returned. The key should be in range(maxkey). |
|
232 |
""" |
|
233 |
s = bprime + int(ceil(log2(log2(aprime)))) |
|
234 |
p = nextprime(2**(2*s)) |
|
235 |
len_f = int(ceil(log2(ceil(aprime/s)))) or 1 |
|
236 |
maxkey = ((p-1)*p)**len_f |
|
237 |
if key is None: |
|
238 |
return maxkey |
|
239 |
keys = [] |
|
240 |
for thiskey in int2list(key, xmax=maxkey, s=(p-1)*p): |
|
241 |
q, r = divmod(thiskey, p) |
|
242 |
q += 1 |
|
243 |
keys.append( (q,r) ) |
|
244 |
def fprime(m): |
|
245 |
"This is a hash function returned by Hprime_H1_compact()." |
|
246 |
substrings = int2list(m, xmax=2**aprime, s=2**(2*s)) |
|
247 |
for q,r in keys: |
|
248 |
for i in xrange(len(substrings)//2): |
|
249 |
substrings[i:i+2] = [(((q*substrings[i ]+r)%p)%(2**s)) + \ |
|
250 |
(((q*substrings[i+1]+r)%p)%(2**s)) * 2**s] |
|
251 |
if len(substrings)%2: |
|
252 |
substrings[-1] = (((q*substrings[ -1]+r)%p)%(2**s)) |
|
253 |
return substrings[0] % 2**bprime |
|
254 |
return fprime |