2014-06-20 20:00:00 +01:00
|
|
|
/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* This package is an SSL implementation written
|
|
|
|
* by Eric Young (eay@cryptsoft.com).
|
|
|
|
* The implementation was written so as to conform with Netscapes SSL.
|
|
|
|
*
|
|
|
|
* This library is free for commercial and non-commercial use as long as
|
|
|
|
* the following conditions are aheared to. The following conditions
|
|
|
|
* apply to all code found in this distribution, be it the RC4, RSA,
|
|
|
|
* lhash, DES, etc., code; not just the SSL code. The SSL documentation
|
|
|
|
* included with this distribution is covered by the same copyright terms
|
|
|
|
* except that the holder is Tim Hudson (tjh@cryptsoft.com).
|
|
|
|
*
|
|
|
|
* Copyright remains Eric Young's, and as such any Copyright notices in
|
|
|
|
* the code are not to be removed.
|
|
|
|
* If this package is used in a product, Eric Young should be given attribution
|
|
|
|
* as the author of the parts of the library used.
|
|
|
|
* This can be in the form of a textual message at program startup or
|
|
|
|
* in documentation (online or textual) provided with the package.
|
|
|
|
*
|
|
|
|
* Redistribution and use in source and binary forms, with or without
|
|
|
|
* modification, are permitted provided that the following conditions
|
|
|
|
* are met:
|
|
|
|
* 1. Redistributions of source code must retain the copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer.
|
|
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
|
|
* documentation and/or other materials provided with the distribution.
|
|
|
|
* 3. All advertising materials mentioning features or use of this software
|
|
|
|
* must display the following acknowledgement:
|
|
|
|
* "This product includes cryptographic software written by
|
|
|
|
* Eric Young (eay@cryptsoft.com)"
|
|
|
|
* The word 'cryptographic' can be left out if the rouines from the library
|
|
|
|
* being used are not cryptographic related :-).
|
|
|
|
* 4. If you include any Windows specific code (or a derivative thereof) from
|
|
|
|
* the apps directory (application code) you must include an acknowledgement:
|
|
|
|
* "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
|
|
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
|
|
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
|
|
* SUCH DAMAGE.
|
|
|
|
*
|
|
|
|
* The licence and distribution terms for any publically available version or
|
|
|
|
* derivative of this code cannot be changed. i.e. this code cannot simply be
|
|
|
|
* copied and put under another distribution licence
|
|
|
|
* [including the GNU Public Licence.] */
|
|
|
|
|
|
|
|
#include <openssl/bn.h>
|
|
|
|
|
Verify RSA private key operation regardless of whether CRT is used.
Previously, the verification was only done when using the CRT method,
as the CRT method has been shown to be extremely sensitive to fault
attacks. However, there's no reason to avoid doing the verification
when the non-CRT method is used (performance-sensitive applications
should always be using the CRT-capable keys).
Previously, when we detected a fault (attack) through this verification,
libcrypto would fall back to the non-CRT method and assume that the
non-CRT method would give a correct result, despite having just
detecting corruption that is likely from an attack. Instead, just give
up, like NSS does.
Previously, the code tried to handle the case where the input was not
reduced mod rsa->n. This is (was) not possible, so avoid trying to
handle that. This simplifies the equality check and lets us use
|CRYPTO_memcmp|.
Change-Id: I78d1e55520a1c8c280cae2b7256e12ff6290507d
Reviewed-on: https://boringssl-review.googlesource.com/7582
Reviewed-by: David Benjamin <davidben@google.com>
2016-03-25 22:23:16 +00:00
|
|
|
#include <openssl/mem.h>
|
2017-11-09 17:31:03 +00:00
|
|
|
#include <openssl/type_check.h>
|
Verify RSA private key operation regardless of whether CRT is used.
Previously, the verification was only done when using the CRT method,
as the CRT method has been shown to be extremely sensitive to fault
attacks. However, there's no reason to avoid doing the verification
when the non-CRT method is used (performance-sensitive applications
should always be using the CRT-capable keys).
Previously, when we detected a fault (attack) through this verification,
libcrypto would fall back to the non-CRT method and assume that the
non-CRT method would give a correct result, despite having just
detecting corruption that is likely from an attack. Instead, just give
up, like NSS does.
Previously, the code tried to handle the case where the input was not
reduced mod rsa->n. This is (was) not possible, so avoid trying to
handle that. This simplifies the equality check and lets us use
|CRYPTO_memcmp|.
Change-Id: I78d1e55520a1c8c280cae2b7256e12ff6290507d
Reviewed-on: https://boringssl-review.googlesource.com/7582
Reviewed-by: David Benjamin <davidben@google.com>
2016-03-25 22:23:16 +00:00
|
|
|
|
2014-06-20 20:00:00 +01:00
|
|
|
#include "internal.h"
|
2017-11-09 17:31:03 +00:00
|
|
|
#include "../../internal.h"
|
2014-06-20 20:00:00 +01:00
|
|
|
|
|
|
|
|
2018-01-27 23:54:28 +00:00
|
|
|
static int bn_cmp_words_consttime(const BN_ULONG *a, size_t a_len,
|
|
|
|
const BN_ULONG *b, size_t b_len) {
|
2018-10-24 23:08:00 +01:00
|
|
|
OPENSSL_STATIC_ASSERT(sizeof(BN_ULONG) <= sizeof(crypto_word_t),
|
|
|
|
"crypto_word_t is too small");
|
2018-01-27 23:54:28 +00:00
|
|
|
int ret = 0;
|
|
|
|
// Process the common words in little-endian order.
|
|
|
|
size_t min = a_len < b_len ? a_len : b_len;
|
|
|
|
for (size_t i = 0; i < min; i++) {
|
|
|
|
crypto_word_t eq = constant_time_eq_w(a[i], b[i]);
|
|
|
|
crypto_word_t lt = constant_time_lt_w(a[i], b[i]);
|
|
|
|
ret =
|
|
|
|
constant_time_select_int(eq, ret, constant_time_select_int(lt, -1, 1));
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2018-01-27 23:54:28 +00:00
|
|
|
// If |a| or |b| has non-zero words beyond |min|, they take precedence.
|
|
|
|
if (a_len < b_len) {
|
|
|
|
crypto_word_t mask = 0;
|
|
|
|
for (size_t i = a_len; i < b_len; i++) {
|
|
|
|
mask |= b[i];
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
2018-01-27 23:54:28 +00:00
|
|
|
ret = constant_time_select_int(constant_time_is_zero_w(mask), ret, -1);
|
|
|
|
} else if (b_len < a_len) {
|
|
|
|
crypto_word_t mask = 0;
|
|
|
|
for (size_t i = b_len; i < a_len; i++) {
|
|
|
|
mask |= a[i];
|
|
|
|
}
|
|
|
|
ret = constant_time_select_int(constant_time_is_zero_w(mask), ret, 1);
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2018-01-27 23:54:28 +00:00
|
|
|
return ret;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2018-01-27 23:54:28 +00:00
|
|
|
int BN_ucmp(const BIGNUM *a, const BIGNUM *b) {
|
|
|
|
return bn_cmp_words_consttime(a->d, a->width, b->d, b->width);
|
|
|
|
}
|
2014-06-20 20:00:00 +01:00
|
|
|
|
2018-01-27 23:54:28 +00:00
|
|
|
int BN_cmp(const BIGNUM *a, const BIGNUM *b) {
|
2014-06-20 20:00:00 +01:00
|
|
|
if ((a == NULL) || (b == NULL)) {
|
|
|
|
if (a != NULL) {
|
|
|
|
return -1;
|
|
|
|
} else if (b != NULL) {
|
|
|
|
return 1;
|
|
|
|
} else {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-01-27 23:54:28 +00:00
|
|
|
// We do not attempt to process the sign bit in constant time. Negative
|
|
|
|
// |BIGNUM|s should never occur in crypto, only calculators.
|
2014-06-20 20:00:00 +01:00
|
|
|
if (a->neg != b->neg) {
|
|
|
|
if (a->neg) {
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
return 1;
|
|
|
|
}
|
2018-01-20 23:55:47 +00:00
|
|
|
|
2018-01-27 23:54:28 +00:00
|
|
|
int ret = BN_ucmp(a, b);
|
|
|
|
return a->neg ? -ret : ret;
|
2017-11-09 17:31:03 +00:00
|
|
|
}
|
|
|
|
|
2018-01-20 23:55:47 +00:00
|
|
|
int bn_less_than_words(const BN_ULONG *a, const BN_ULONG *b, size_t len) {
|
2018-01-27 23:54:28 +00:00
|
|
|
return bn_cmp_words_consttime(a, len, b, len) < 0;
|
2018-01-20 23:55:47 +00:00
|
|
|
}
|
|
|
|
|
2014-06-20 20:00:00 +01:00
|
|
|
int BN_abs_is_word(const BIGNUM *bn, BN_ULONG w) {
|
2018-02-04 00:54:36 +00:00
|
|
|
if (bn->width == 0) {
|
|
|
|
return w == 0;
|
|
|
|
}
|
|
|
|
BN_ULONG mask = bn->d[0] ^ w;
|
|
|
|
for (int i = 1; i < bn->width; i++) {
|
|
|
|
mask |= bn->d[i];
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
2018-02-04 00:54:36 +00:00
|
|
|
return mask == 0;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
2016-07-25 21:36:58 +01:00
|
|
|
int BN_cmp_word(const BIGNUM *a, BN_ULONG b) {
|
|
|
|
BIGNUM b_bn;
|
|
|
|
BN_init(&b_bn);
|
|
|
|
|
|
|
|
b_bn.d = &b;
|
2018-01-15 10:23:24 +00:00
|
|
|
b_bn.width = b > 0;
|
2016-07-25 21:36:58 +01:00
|
|
|
b_bn.dmax = 1;
|
|
|
|
b_bn.flags = BN_FLG_STATIC_DATA;
|
|
|
|
return BN_cmp(a, &b_bn);
|
|
|
|
}
|
|
|
|
|
2014-06-20 20:00:00 +01:00
|
|
|
int BN_is_zero(const BIGNUM *bn) {
|
2018-02-04 00:54:36 +00:00
|
|
|
return bn_fits_in_words(bn, 0);
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
int BN_is_one(const BIGNUM *bn) {
|
|
|
|
return bn->neg == 0 && BN_abs_is_word(bn, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
int BN_is_word(const BIGNUM *bn, BN_ULONG w) {
|
|
|
|
return BN_abs_is_word(bn, w) && (w == 0 || bn->neg == 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
int BN_is_odd(const BIGNUM *bn) {
|
2018-01-15 10:23:24 +00:00
|
|
|
return bn->width > 0 && (bn->d[0] & 1) == 1;
|
2014-06-20 20:00:00 +01:00
|
|
|
}
|
Verify RSA private key operation regardless of whether CRT is used.
Previously, the verification was only done when using the CRT method,
as the CRT method has been shown to be extremely sensitive to fault
attacks. However, there's no reason to avoid doing the verification
when the non-CRT method is used (performance-sensitive applications
should always be using the CRT-capable keys).
Previously, when we detected a fault (attack) through this verification,
libcrypto would fall back to the non-CRT method and assume that the
non-CRT method would give a correct result, despite having just
detecting corruption that is likely from an attack. Instead, just give
up, like NSS does.
Previously, the code tried to handle the case where the input was not
reduced mod rsa->n. This is (was) not possible, so avoid trying to
handle that. This simplifies the equality check and lets us use
|CRYPTO_memcmp|.
Change-Id: I78d1e55520a1c8c280cae2b7256e12ff6290507d
Reviewed-on: https://boringssl-review.googlesource.com/7582
Reviewed-by: David Benjamin <davidben@google.com>
2016-03-25 22:23:16 +00:00
|
|
|
|
2017-01-30 19:52:27 +00:00
|
|
|
int BN_is_pow2(const BIGNUM *bn) {
|
Add initial support for non-minimal BIGNUMs.
Thanks to Andres Erbsen for extremely helpful suggestions on how finally
plug this long-standing hole!
OpenSSL BIGNUMs are currently minimal-width, which means they cannot be
constant-time. We'll need to either excise BIGNUM from RSA and EC or
somehow fix BIGNUM. EC_SCALAR and later EC_FELEM work will excise it
from EC, but RSA's BIGNUMs are more transparent. Teaching BIGNUM to
handle non-minimal word widths is probably simpler.
The main constraint is BIGNUM's large "calculator" API surface. One
could, in theory, do arbitrary math on RSA components, which means all
public functions must tolerate non-minimal inputs. This is also useful
for EC; https://boringssl-review.googlesource.com/c/boringssl/+/24445 is
silly.
As a first step, fix comparison-type functions that were assuming
minimal BIGNUMs. I've also added bn_resize_words, but it is testing-only
until the rest of the library is fixed.
bn->top is now a loose upper bound we carry around. It does not affect
numerical results, only performance and secrecy. This is a departure
from the original meaning, and compiler help in auditing everything is
nice, so the final change in this series will rename bn->top to
bn->width. Thus these new functions are named per "width", not "top".
Looking further ahead, how are output BIGNUM widths determined? There's
three notions of correctness here:
1. Do I compute the right answer for all widths?
2. Do I handle secret data in constant time?
3. Does my memory usage not balloon absurdly?
For (1), a BIGNUM function must give the same answer for all input
widths. BN_mod_add_quick may assume |a| < |m|, but |a| may still be
wider than |m| by way of leading zeres. The simplest approach is to
write code in a width-agnostic way and rely on functions to accept all
widths. Where functions need to look at bn->d, we'll a few helper
functions to smooth over funny widths.
For (2), (1) is little cumbersome. Consider constant-time modular
addition. A sane type system would guarantee input widths match. But C
is weak here, and bifurcating the internals is a lot of work. Thus, at
least for now, I do not propose we move RSA's internal computation out
of BIGNUM. (EC_SCALAR/EC_FELEM are valuable for EC because we get to
stack-allocate, curves were already specialized, and EC only has two
types with many operations on those types. None of these apply to RSA.
We've got numbers mod n, mod p, mod q, and their corresponding
exponents, each of which is used for basically one operation.)
Instead, constant-time BIGNUM functions will output non-minimal widths.
This is trivial for BN_bin2bn or modular arithmetic. But for BN_mul,
constant-time[*] would dictate r->top = a->top + b->top. A calculator
repeatedly multiplying by one would then run out of memory. Those we'll
split into a private BN_mul_fixed for crypto, leaving BN_mul for
calculators. BN_mul is just BN_mul_fixed followed by bn_correct_top.
[*] BN_mul is not constant-time for other reasons, but that will be
fixed separately.
Bug: 232
Change-Id: Ide2258ae8c09a9a41bb71d6777908d1c27917069
Reviewed-on: https://boringssl-review.googlesource.com/25244
Reviewed-by: Adam Langley <agl@google.com>
2018-01-20 20:56:53 +00:00
|
|
|
int width = bn_minimal_width(bn);
|
|
|
|
if (width == 0 || bn->neg) {
|
2017-01-30 19:52:27 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
Add initial support for non-minimal BIGNUMs.
Thanks to Andres Erbsen for extremely helpful suggestions on how finally
plug this long-standing hole!
OpenSSL BIGNUMs are currently minimal-width, which means they cannot be
constant-time. We'll need to either excise BIGNUM from RSA and EC or
somehow fix BIGNUM. EC_SCALAR and later EC_FELEM work will excise it
from EC, but RSA's BIGNUMs are more transparent. Teaching BIGNUM to
handle non-minimal word widths is probably simpler.
The main constraint is BIGNUM's large "calculator" API surface. One
could, in theory, do arbitrary math on RSA components, which means all
public functions must tolerate non-minimal inputs. This is also useful
for EC; https://boringssl-review.googlesource.com/c/boringssl/+/24445 is
silly.
As a first step, fix comparison-type functions that were assuming
minimal BIGNUMs. I've also added bn_resize_words, but it is testing-only
until the rest of the library is fixed.
bn->top is now a loose upper bound we carry around. It does not affect
numerical results, only performance and secrecy. This is a departure
from the original meaning, and compiler help in auditing everything is
nice, so the final change in this series will rename bn->top to
bn->width. Thus these new functions are named per "width", not "top".
Looking further ahead, how are output BIGNUM widths determined? There's
three notions of correctness here:
1. Do I compute the right answer for all widths?
2. Do I handle secret data in constant time?
3. Does my memory usage not balloon absurdly?
For (1), a BIGNUM function must give the same answer for all input
widths. BN_mod_add_quick may assume |a| < |m|, but |a| may still be
wider than |m| by way of leading zeres. The simplest approach is to
write code in a width-agnostic way and rely on functions to accept all
widths. Where functions need to look at bn->d, we'll a few helper
functions to smooth over funny widths.
For (2), (1) is little cumbersome. Consider constant-time modular
addition. A sane type system would guarantee input widths match. But C
is weak here, and bifurcating the internals is a lot of work. Thus, at
least for now, I do not propose we move RSA's internal computation out
of BIGNUM. (EC_SCALAR/EC_FELEM are valuable for EC because we get to
stack-allocate, curves were already specialized, and EC only has two
types with many operations on those types. None of these apply to RSA.
We've got numbers mod n, mod p, mod q, and their corresponding
exponents, each of which is used for basically one operation.)
Instead, constant-time BIGNUM functions will output non-minimal widths.
This is trivial for BN_bin2bn or modular arithmetic. But for BN_mul,
constant-time[*] would dictate r->top = a->top + b->top. A calculator
repeatedly multiplying by one would then run out of memory. Those we'll
split into a private BN_mul_fixed for crypto, leaving BN_mul for
calculators. BN_mul is just BN_mul_fixed followed by bn_correct_top.
[*] BN_mul is not constant-time for other reasons, but that will be
fixed separately.
Bug: 232
Change-Id: Ide2258ae8c09a9a41bb71d6777908d1c27917069
Reviewed-on: https://boringssl-review.googlesource.com/25244
Reviewed-by: Adam Langley <agl@google.com>
2018-01-20 20:56:53 +00:00
|
|
|
for (int i = 0; i < width - 1; i++) {
|
2017-01-30 19:52:27 +00:00
|
|
|
if (bn->d[i] != 0) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Add initial support for non-minimal BIGNUMs.
Thanks to Andres Erbsen for extremely helpful suggestions on how finally
plug this long-standing hole!
OpenSSL BIGNUMs are currently minimal-width, which means they cannot be
constant-time. We'll need to either excise BIGNUM from RSA and EC or
somehow fix BIGNUM. EC_SCALAR and later EC_FELEM work will excise it
from EC, but RSA's BIGNUMs are more transparent. Teaching BIGNUM to
handle non-minimal word widths is probably simpler.
The main constraint is BIGNUM's large "calculator" API surface. One
could, in theory, do arbitrary math on RSA components, which means all
public functions must tolerate non-minimal inputs. This is also useful
for EC; https://boringssl-review.googlesource.com/c/boringssl/+/24445 is
silly.
As a first step, fix comparison-type functions that were assuming
minimal BIGNUMs. I've also added bn_resize_words, but it is testing-only
until the rest of the library is fixed.
bn->top is now a loose upper bound we carry around. It does not affect
numerical results, only performance and secrecy. This is a departure
from the original meaning, and compiler help in auditing everything is
nice, so the final change in this series will rename bn->top to
bn->width. Thus these new functions are named per "width", not "top".
Looking further ahead, how are output BIGNUM widths determined? There's
three notions of correctness here:
1. Do I compute the right answer for all widths?
2. Do I handle secret data in constant time?
3. Does my memory usage not balloon absurdly?
For (1), a BIGNUM function must give the same answer for all input
widths. BN_mod_add_quick may assume |a| < |m|, but |a| may still be
wider than |m| by way of leading zeres. The simplest approach is to
write code in a width-agnostic way and rely on functions to accept all
widths. Where functions need to look at bn->d, we'll a few helper
functions to smooth over funny widths.
For (2), (1) is little cumbersome. Consider constant-time modular
addition. A sane type system would guarantee input widths match. But C
is weak here, and bifurcating the internals is a lot of work. Thus, at
least for now, I do not propose we move RSA's internal computation out
of BIGNUM. (EC_SCALAR/EC_FELEM are valuable for EC because we get to
stack-allocate, curves were already specialized, and EC only has two
types with many operations on those types. None of these apply to RSA.
We've got numbers mod n, mod p, mod q, and their corresponding
exponents, each of which is used for basically one operation.)
Instead, constant-time BIGNUM functions will output non-minimal widths.
This is trivial for BN_bin2bn or modular arithmetic. But for BN_mul,
constant-time[*] would dictate r->top = a->top + b->top. A calculator
repeatedly multiplying by one would then run out of memory. Those we'll
split into a private BN_mul_fixed for crypto, leaving BN_mul for
calculators. BN_mul is just BN_mul_fixed followed by bn_correct_top.
[*] BN_mul is not constant-time for other reasons, but that will be
fixed separately.
Bug: 232
Change-Id: Ide2258ae8c09a9a41bb71d6777908d1c27917069
Reviewed-on: https://boringssl-review.googlesource.com/25244
Reviewed-by: Adam Langley <agl@google.com>
2018-01-20 20:56:53 +00:00
|
|
|
return 0 == (bn->d[width-1] & (bn->d[width-1] - 1));
|
2017-01-30 19:52:27 +00:00
|
|
|
}
|
|
|
|
|
Verify RSA private key operation regardless of whether CRT is used.
Previously, the verification was only done when using the CRT method,
as the CRT method has been shown to be extremely sensitive to fault
attacks. However, there's no reason to avoid doing the verification
when the non-CRT method is used (performance-sensitive applications
should always be using the CRT-capable keys).
Previously, when we detected a fault (attack) through this verification,
libcrypto would fall back to the non-CRT method and assume that the
non-CRT method would give a correct result, despite having just
detecting corruption that is likely from an attack. Instead, just give
up, like NSS does.
Previously, the code tried to handle the case where the input was not
reduced mod rsa->n. This is (was) not possible, so avoid trying to
handle that. This simplifies the equality check and lets us use
|CRYPTO_memcmp|.
Change-Id: I78d1e55520a1c8c280cae2b7256e12ff6290507d
Reviewed-on: https://boringssl-review.googlesource.com/7582
Reviewed-by: David Benjamin <davidben@google.com>
2016-03-25 22:23:16 +00:00
|
|
|
int BN_equal_consttime(const BIGNUM *a, const BIGNUM *b) {
|
Add initial support for non-minimal BIGNUMs.
Thanks to Andres Erbsen for extremely helpful suggestions on how finally
plug this long-standing hole!
OpenSSL BIGNUMs are currently minimal-width, which means they cannot be
constant-time. We'll need to either excise BIGNUM from RSA and EC or
somehow fix BIGNUM. EC_SCALAR and later EC_FELEM work will excise it
from EC, but RSA's BIGNUMs are more transparent. Teaching BIGNUM to
handle non-minimal word widths is probably simpler.
The main constraint is BIGNUM's large "calculator" API surface. One
could, in theory, do arbitrary math on RSA components, which means all
public functions must tolerate non-minimal inputs. This is also useful
for EC; https://boringssl-review.googlesource.com/c/boringssl/+/24445 is
silly.
As a first step, fix comparison-type functions that were assuming
minimal BIGNUMs. I've also added bn_resize_words, but it is testing-only
until the rest of the library is fixed.
bn->top is now a loose upper bound we carry around. It does not affect
numerical results, only performance and secrecy. This is a departure
from the original meaning, and compiler help in auditing everything is
nice, so the final change in this series will rename bn->top to
bn->width. Thus these new functions are named per "width", not "top".
Looking further ahead, how are output BIGNUM widths determined? There's
three notions of correctness here:
1. Do I compute the right answer for all widths?
2. Do I handle secret data in constant time?
3. Does my memory usage not balloon absurdly?
For (1), a BIGNUM function must give the same answer for all input
widths. BN_mod_add_quick may assume |a| < |m|, but |a| may still be
wider than |m| by way of leading zeres. The simplest approach is to
write code in a width-agnostic way and rely on functions to accept all
widths. Where functions need to look at bn->d, we'll a few helper
functions to smooth over funny widths.
For (2), (1) is little cumbersome. Consider constant-time modular
addition. A sane type system would guarantee input widths match. But C
is weak here, and bifurcating the internals is a lot of work. Thus, at
least for now, I do not propose we move RSA's internal computation out
of BIGNUM. (EC_SCALAR/EC_FELEM are valuable for EC because we get to
stack-allocate, curves were already specialized, and EC only has two
types with many operations on those types. None of these apply to RSA.
We've got numbers mod n, mod p, mod q, and their corresponding
exponents, each of which is used for basically one operation.)
Instead, constant-time BIGNUM functions will output non-minimal widths.
This is trivial for BN_bin2bn or modular arithmetic. But for BN_mul,
constant-time[*] would dictate r->top = a->top + b->top. A calculator
repeatedly multiplying by one would then run out of memory. Those we'll
split into a private BN_mul_fixed for crypto, leaving BN_mul for
calculators. BN_mul is just BN_mul_fixed followed by bn_correct_top.
[*] BN_mul is not constant-time for other reasons, but that will be
fixed separately.
Bug: 232
Change-Id: Ide2258ae8c09a9a41bb71d6777908d1c27917069
Reviewed-on: https://boringssl-review.googlesource.com/25244
Reviewed-by: Adam Langley <agl@google.com>
2018-01-20 20:56:53 +00:00
|
|
|
BN_ULONG mask = 0;
|
|
|
|
// If |a| or |b| has more words than the other, all those words must be zero.
|
2018-01-15 10:23:24 +00:00
|
|
|
for (int i = a->width; i < b->width; i++) {
|
Add initial support for non-minimal BIGNUMs.
Thanks to Andres Erbsen for extremely helpful suggestions on how finally
plug this long-standing hole!
OpenSSL BIGNUMs are currently minimal-width, which means they cannot be
constant-time. We'll need to either excise BIGNUM from RSA and EC or
somehow fix BIGNUM. EC_SCALAR and later EC_FELEM work will excise it
from EC, but RSA's BIGNUMs are more transparent. Teaching BIGNUM to
handle non-minimal word widths is probably simpler.
The main constraint is BIGNUM's large "calculator" API surface. One
could, in theory, do arbitrary math on RSA components, which means all
public functions must tolerate non-minimal inputs. This is also useful
for EC; https://boringssl-review.googlesource.com/c/boringssl/+/24445 is
silly.
As a first step, fix comparison-type functions that were assuming
minimal BIGNUMs. I've also added bn_resize_words, but it is testing-only
until the rest of the library is fixed.
bn->top is now a loose upper bound we carry around. It does not affect
numerical results, only performance and secrecy. This is a departure
from the original meaning, and compiler help in auditing everything is
nice, so the final change in this series will rename bn->top to
bn->width. Thus these new functions are named per "width", not "top".
Looking further ahead, how are output BIGNUM widths determined? There's
three notions of correctness here:
1. Do I compute the right answer for all widths?
2. Do I handle secret data in constant time?
3. Does my memory usage not balloon absurdly?
For (1), a BIGNUM function must give the same answer for all input
widths. BN_mod_add_quick may assume |a| < |m|, but |a| may still be
wider than |m| by way of leading zeres. The simplest approach is to
write code in a width-agnostic way and rely on functions to accept all
widths. Where functions need to look at bn->d, we'll a few helper
functions to smooth over funny widths.
For (2), (1) is little cumbersome. Consider constant-time modular
addition. A sane type system would guarantee input widths match. But C
is weak here, and bifurcating the internals is a lot of work. Thus, at
least for now, I do not propose we move RSA's internal computation out
of BIGNUM. (EC_SCALAR/EC_FELEM are valuable for EC because we get to
stack-allocate, curves were already specialized, and EC only has two
types with many operations on those types. None of these apply to RSA.
We've got numbers mod n, mod p, mod q, and their corresponding
exponents, each of which is used for basically one operation.)
Instead, constant-time BIGNUM functions will output non-minimal widths.
This is trivial for BN_bin2bn or modular arithmetic. But for BN_mul,
constant-time[*] would dictate r->top = a->top + b->top. A calculator
repeatedly multiplying by one would then run out of memory. Those we'll
split into a private BN_mul_fixed for crypto, leaving BN_mul for
calculators. BN_mul is just BN_mul_fixed followed by bn_correct_top.
[*] BN_mul is not constant-time for other reasons, but that will be
fixed separately.
Bug: 232
Change-Id: Ide2258ae8c09a9a41bb71d6777908d1c27917069
Reviewed-on: https://boringssl-review.googlesource.com/25244
Reviewed-by: Adam Langley <agl@google.com>
2018-01-20 20:56:53 +00:00
|
|
|
mask |= b->d[i];
|
Verify RSA private key operation regardless of whether CRT is used.
Previously, the verification was only done when using the CRT method,
as the CRT method has been shown to be extremely sensitive to fault
attacks. However, there's no reason to avoid doing the verification
when the non-CRT method is used (performance-sensitive applications
should always be using the CRT-capable keys).
Previously, when we detected a fault (attack) through this verification,
libcrypto would fall back to the non-CRT method and assume that the
non-CRT method would give a correct result, despite having just
detecting corruption that is likely from an attack. Instead, just give
up, like NSS does.
Previously, the code tried to handle the case where the input was not
reduced mod rsa->n. This is (was) not possible, so avoid trying to
handle that. This simplifies the equality check and lets us use
|CRYPTO_memcmp|.
Change-Id: I78d1e55520a1c8c280cae2b7256e12ff6290507d
Reviewed-on: https://boringssl-review.googlesource.com/7582
Reviewed-by: David Benjamin <davidben@google.com>
2016-03-25 22:23:16 +00:00
|
|
|
}
|
2018-01-15 10:23:24 +00:00
|
|
|
for (int i = b->width; i < a->width; i++) {
|
Add initial support for non-minimal BIGNUMs.
Thanks to Andres Erbsen for extremely helpful suggestions on how finally
plug this long-standing hole!
OpenSSL BIGNUMs are currently minimal-width, which means they cannot be
constant-time. We'll need to either excise BIGNUM from RSA and EC or
somehow fix BIGNUM. EC_SCALAR and later EC_FELEM work will excise it
from EC, but RSA's BIGNUMs are more transparent. Teaching BIGNUM to
handle non-minimal word widths is probably simpler.
The main constraint is BIGNUM's large "calculator" API surface. One
could, in theory, do arbitrary math on RSA components, which means all
public functions must tolerate non-minimal inputs. This is also useful
for EC; https://boringssl-review.googlesource.com/c/boringssl/+/24445 is
silly.
As a first step, fix comparison-type functions that were assuming
minimal BIGNUMs. I've also added bn_resize_words, but it is testing-only
until the rest of the library is fixed.
bn->top is now a loose upper bound we carry around. It does not affect
numerical results, only performance and secrecy. This is a departure
from the original meaning, and compiler help in auditing everything is
nice, so the final change in this series will rename bn->top to
bn->width. Thus these new functions are named per "width", not "top".
Looking further ahead, how are output BIGNUM widths determined? There's
three notions of correctness here:
1. Do I compute the right answer for all widths?
2. Do I handle secret data in constant time?
3. Does my memory usage not balloon absurdly?
For (1), a BIGNUM function must give the same answer for all input
widths. BN_mod_add_quick may assume |a| < |m|, but |a| may still be
wider than |m| by way of leading zeres. The simplest approach is to
write code in a width-agnostic way and rely on functions to accept all
widths. Where functions need to look at bn->d, we'll a few helper
functions to smooth over funny widths.
For (2), (1) is little cumbersome. Consider constant-time modular
addition. A sane type system would guarantee input widths match. But C
is weak here, and bifurcating the internals is a lot of work. Thus, at
least for now, I do not propose we move RSA's internal computation out
of BIGNUM. (EC_SCALAR/EC_FELEM are valuable for EC because we get to
stack-allocate, curves were already specialized, and EC only has two
types with many operations on those types. None of these apply to RSA.
We've got numbers mod n, mod p, mod q, and their corresponding
exponents, each of which is used for basically one operation.)
Instead, constant-time BIGNUM functions will output non-minimal widths.
This is trivial for BN_bin2bn or modular arithmetic. But for BN_mul,
constant-time[*] would dictate r->top = a->top + b->top. A calculator
repeatedly multiplying by one would then run out of memory. Those we'll
split into a private BN_mul_fixed for crypto, leaving BN_mul for
calculators. BN_mul is just BN_mul_fixed followed by bn_correct_top.
[*] BN_mul is not constant-time for other reasons, but that will be
fixed separately.
Bug: 232
Change-Id: Ide2258ae8c09a9a41bb71d6777908d1c27917069
Reviewed-on: https://boringssl-review.googlesource.com/25244
Reviewed-by: Adam Langley <agl@google.com>
2018-01-20 20:56:53 +00:00
|
|
|
mask |= a->d[i];
|
|
|
|
}
|
|
|
|
// Common words must match.
|
2018-01-15 10:23:24 +00:00
|
|
|
int min = a->width < b->width ? a->width : b->width;
|
Add initial support for non-minimal BIGNUMs.
Thanks to Andres Erbsen for extremely helpful suggestions on how finally
plug this long-standing hole!
OpenSSL BIGNUMs are currently minimal-width, which means they cannot be
constant-time. We'll need to either excise BIGNUM from RSA and EC or
somehow fix BIGNUM. EC_SCALAR and later EC_FELEM work will excise it
from EC, but RSA's BIGNUMs are more transparent. Teaching BIGNUM to
handle non-minimal word widths is probably simpler.
The main constraint is BIGNUM's large "calculator" API surface. One
could, in theory, do arbitrary math on RSA components, which means all
public functions must tolerate non-minimal inputs. This is also useful
for EC; https://boringssl-review.googlesource.com/c/boringssl/+/24445 is
silly.
As a first step, fix comparison-type functions that were assuming
minimal BIGNUMs. I've also added bn_resize_words, but it is testing-only
until the rest of the library is fixed.
bn->top is now a loose upper bound we carry around. It does not affect
numerical results, only performance and secrecy. This is a departure
from the original meaning, and compiler help in auditing everything is
nice, so the final change in this series will rename bn->top to
bn->width. Thus these new functions are named per "width", not "top".
Looking further ahead, how are output BIGNUM widths determined? There's
three notions of correctness here:
1. Do I compute the right answer for all widths?
2. Do I handle secret data in constant time?
3. Does my memory usage not balloon absurdly?
For (1), a BIGNUM function must give the same answer for all input
widths. BN_mod_add_quick may assume |a| < |m|, but |a| may still be
wider than |m| by way of leading zeres. The simplest approach is to
write code in a width-agnostic way and rely on functions to accept all
widths. Where functions need to look at bn->d, we'll a few helper
functions to smooth over funny widths.
For (2), (1) is little cumbersome. Consider constant-time modular
addition. A sane type system would guarantee input widths match. But C
is weak here, and bifurcating the internals is a lot of work. Thus, at
least for now, I do not propose we move RSA's internal computation out
of BIGNUM. (EC_SCALAR/EC_FELEM are valuable for EC because we get to
stack-allocate, curves were already specialized, and EC only has two
types with many operations on those types. None of these apply to RSA.
We've got numbers mod n, mod p, mod q, and their corresponding
exponents, each of which is used for basically one operation.)
Instead, constant-time BIGNUM functions will output non-minimal widths.
This is trivial for BN_bin2bn or modular arithmetic. But for BN_mul,
constant-time[*] would dictate r->top = a->top + b->top. A calculator
repeatedly multiplying by one would then run out of memory. Those we'll
split into a private BN_mul_fixed for crypto, leaving BN_mul for
calculators. BN_mul is just BN_mul_fixed followed by bn_correct_top.
[*] BN_mul is not constant-time for other reasons, but that will be
fixed separately.
Bug: 232
Change-Id: Ide2258ae8c09a9a41bb71d6777908d1c27917069
Reviewed-on: https://boringssl-review.googlesource.com/25244
Reviewed-by: Adam Langley <agl@google.com>
2018-01-20 20:56:53 +00:00
|
|
|
for (int i = 0; i < min; i++) {
|
|
|
|
mask |= (a->d[i] ^ b->d[i]);
|
|
|
|
}
|
|
|
|
// The sign bit must match.
|
|
|
|
mask |= (a->neg ^ b->neg);
|
|
|
|
return mask == 0;
|
Verify RSA private key operation regardless of whether CRT is used.
Previously, the verification was only done when using the CRT method,
as the CRT method has been shown to be extremely sensitive to fault
attacks. However, there's no reason to avoid doing the verification
when the non-CRT method is used (performance-sensitive applications
should always be using the CRT-capable keys).
Previously, when we detected a fault (attack) through this verification,
libcrypto would fall back to the non-CRT method and assume that the
non-CRT method would give a correct result, despite having just
detecting corruption that is likely from an attack. Instead, just give
up, like NSS does.
Previously, the code tried to handle the case where the input was not
reduced mod rsa->n. This is (was) not possible, so avoid trying to
handle that. This simplifies the equality check and lets us use
|CRYPTO_memcmp|.
Change-Id: I78d1e55520a1c8c280cae2b7256e12ff6290507d
Reviewed-on: https://boringssl-review.googlesource.com/7582
Reviewed-by: David Benjamin <davidben@google.com>
2016-03-25 22:23:16 +00:00
|
|
|
}
|