Skip to content

Commit b18162a

Browse files
bbbrumleyromen
bbbrumley
authored and
romen
committed
CVE-2018-5407 fix: ECC ladder
Reviewed-by: Matt Caswell <matt@openssl.org> Reviewed-by: Paul Dale <paul.dale@oracle.com> Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com> (Merged from #7593)
1 parent 59b9c67 commit b18162a

File tree

3 files changed

+291
-0
lines changed

3 files changed

+291
-0
lines changed

‎CHANGES‎

Lines changed: 13 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -9,6 +9,19 @@
99

1010
Changes between 1.0.2p and 1.0.2q [xx XXX xxxx]
1111

12+
*) Microarchitecture timing vulnerability in ECC scalar multiplication
13+
14+
OpenSSL ECC scalar multiplication, used in e.g. ECDSA and ECDH, has been
15+
shown to be vulnerable to a microarchitecture timing side channel attack.
16+
An attacker with sufficient access to mount local timing attacks during
17+
ECDSA signature generation could recover the private key.
18+
19+
This issue was reported to OpenSSL on 26th October 2018 by Alejandro
20+
Cabrera Aldaya, Billy Brumley, Sohaib ul Hassan, Cesar Pereida Garcia and
21+
Nicola Tuveri.
22+
(CVE-2018-5407)
23+
[Billy Brumley]
24+
1225
*) Resolve a compatibility issue in EC_GROUP handling with the FIPS Object
1326
Module, accidentally introduced while backporting security fixes from the
1427
development branch and hindering the use of ECC in FIPS mode.

‎crypto/bn/bn_lib.c‎

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -903,6 +903,38 @@ void BN_consttime_swap(BN_ULONG condition, BIGNUM *a, BIGNUM *b, int nwords)
903903
a->top ^= t;
904904
b->top ^= t;
905905

906+
t = (a->neg ^ b->neg) & condition;
907+
a->neg ^= t;
908+
b->neg ^= t;
909+
910+
/*-
911+
* BN_FLG_STATIC_DATA: indicates that data may not be written to. Intention
912+
* is actually to treat it as it's read-only data, and some (if not most)
913+
* of it does reside in read-only segment. In other words observation of
914+
* BN_FLG_STATIC_DATA in BN_consttime_swap should be treated as fatal
915+
* condition. It would either cause SEGV or effectively cause data
916+
* corruption.
917+
*
918+
* BN_FLG_MALLOCED: refers to BN structure itself, and hence must be
919+
* preserved.
920+
*
921+
* BN_FLG_SECURE: must be preserved, because it determines how x->d was
922+
* allocated and hence how to free it.
923+
*
924+
* BN_FLG_CONSTTIME: sufficient to mask and swap
925+
*
926+
* BN_FLG_FIXED_TOP: indicates that we haven't called bn_correct_top() on
927+
* the data, so the d array may be padded with additional 0 values (i.e.
928+
* top could be greater than the minimal value that it could be). We should
929+
* be swapping it
930+
*/
931+
932+
#define BN_CONSTTIME_SWAP_FLAGS (BN_FLG_CONSTTIME | BN_FLG_FIXED_TOP)
933+
934+
t = ((a->flags ^ b->flags) & BN_CONSTTIME_SWAP_FLAGS) & condition;
935+
a->flags ^= t;
936+
b->flags ^= t;
937+
906938
#define BN_CONSTTIME_SWAP(ind) \
907939
do { \
908940
t = (a->d[ind] ^ b->d[ind]) & condition; \

‎crypto/ec/ec_mult.c‎

Lines changed: 246 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -310,6 +310,224 @@ static signed char *compute_wNAF(const BIGNUM *scalar, int w, size_t *ret_len)
310310
return r;
311311
}
312312

313+
#define EC_POINT_BN_set_flags(P, flags) do { \
314+
BN_set_flags(&(P)->X, (flags)); \
315+
BN_set_flags(&(P)->Y, (flags)); \
316+
BN_set_flags(&(P)->Z, (flags)); \
317+
} while(0)
318+
319+
/*-
320+
* This functions computes (in constant time) a point multiplication over the
321+
* EC group.
322+
*
323+
* At a high level, it is Montgomery ladder with conditional swaps.
324+
*
325+
* It performs either a fixed scalar point multiplication
326+
* (scalar * generator)
327+
* when point is NULL, or a generic scalar point multiplication
328+
* (scalar * point)
329+
* when point is not NULL.
330+
*
331+
* scalar should be in the range [0,n) otherwise all constant time bets are off.
332+
*
333+
* NB: This says nothing about EC_POINT_add and EC_POINT_dbl,
334+
* which of course are not constant time themselves.
335+
*
336+
* The product is stored in r.
337+
*
338+
* Returns 1 on success, 0 otherwise.
339+
*/
340+
static int ec_mul_consttime(const EC_GROUP *group, EC_POINT *r,
341+
const BIGNUM *scalar, const EC_POINT *point,
342+
BN_CTX *ctx)
343+
{
344+
int i, cardinality_bits, group_top, kbit, pbit, Z_is_one;
345+
EC_POINT *s = NULL;
346+
BIGNUM *k = NULL;
347+
BIGNUM *lambda = NULL;
348+
BIGNUM *cardinality = NULL;
349+
BN_CTX *new_ctx = NULL;
350+
int ret = 0;
351+
352+
if (ctx == NULL && (ctx = new_ctx = BN_CTX_new()) == NULL)
353+
return 0;
354+
355+
BN_CTX_start(ctx);
356+
357+
s = EC_POINT_new(group);
358+
if (s == NULL)
359+
goto err;
360+
361+
if (point == NULL) {
362+
if (!EC_POINT_copy(s, group->generator))
363+
goto err;
364+
} else {
365+
if (!EC_POINT_copy(s, point))
366+
goto err;
367+
}
368+
369+
EC_POINT_BN_set_flags(s, BN_FLG_CONSTTIME);
370+
371+
cardinality = BN_CTX_get(ctx);
372+
lambda = BN_CTX_get(ctx);
373+
k = BN_CTX_get(ctx);
374+
if (k == NULL || !BN_mul(cardinality, &group->order, &group->cofactor, ctx))
375+
goto err;
376+
377+
/*
378+
* Group cardinalities are often on a word boundary.
379+
* So when we pad the scalar, some timing diff might
380+
* pop if it needs to be expanded due to carries.
381+
* So expand ahead of time.
382+
*/
383+
cardinality_bits = BN_num_bits(cardinality);
384+
group_top = cardinality->top;
385+
if ((bn_wexpand(k, group_top + 2) == NULL)
386+
|| (bn_wexpand(lambda, group_top + 2) == NULL))
387+
goto err;
388+
389+
if (!BN_copy(k, scalar))
390+
goto err;
391+
392+
BN_set_flags(k, BN_FLG_CONSTTIME);
393+
394+
if ((BN_num_bits(k) > cardinality_bits) || (BN_is_negative(k))) {
395+
/*-
396+
* this is an unusual input, and we don't guarantee
397+
* constant-timeness
398+
*/
399+
if (!BN_nnmod(k, k, cardinality, ctx))
400+
goto err;
401+
}
402+
403+
if (!BN_add(lambda, k, cardinality))
404+
goto err;
405+
BN_set_flags(lambda, BN_FLG_CONSTTIME);
406+
if (!BN_add(k, lambda, cardinality))
407+
goto err;
408+
/*
409+
* lambda := scalar + cardinality
410+
* k := scalar + 2*cardinality
411+
*/
412+
kbit = BN_is_bit_set(lambda, cardinality_bits);
413+
BN_consttime_swap(kbit, k, lambda, group_top + 2);
414+
415+
group_top = group->field.top;
416+
if ((bn_wexpand(&s->X, group_top) == NULL)
417+
|| (bn_wexpand(&s->Y, group_top) == NULL)
418+
|| (bn_wexpand(&s->Z, group_top) == NULL)
419+
|| (bn_wexpand(&r->X, group_top) == NULL)
420+
|| (bn_wexpand(&r->Y, group_top) == NULL)
421+
|| (bn_wexpand(&r->Z, group_top) == NULL))
422+
goto err;
423+
424+
/* top bit is a 1, in a fixed pos */
425+
if (!EC_POINT_copy(r, s))
426+
goto err;
427+
428+
EC_POINT_BN_set_flags(r, BN_FLG_CONSTTIME);
429+
430+
if (!EC_POINT_dbl(group, s, s, ctx))
431+
goto err;
432+
433+
pbit = 0;
434+
435+
#define EC_POINT_CSWAP(c, a, b, w, t) do { \
436+
BN_consttime_swap(c, &(a)->X, &(b)->X, w); \
437+
BN_consttime_swap(c, &(a)->Y, &(b)->Y, w); \
438+
BN_consttime_swap(c, &(a)->Z, &(b)->Z, w); \
439+
t = ((a)->Z_is_one ^ (b)->Z_is_one) & (c); \
440+
(a)->Z_is_one ^= (t); \
441+
(b)->Z_is_one ^= (t); \
442+
} while(0)
443+
444+
/*-
445+
* The ladder step, with branches, is
446+
*
447+
* k[i] == 0: S = add(R, S), R = dbl(R)
448+
* k[i] == 1: R = add(S, R), S = dbl(S)
449+
*
450+
* Swapping R, S conditionally on k[i] leaves you with state
451+
*
452+
* k[i] == 0: T, U = R, S
453+
* k[i] == 1: T, U = S, R
454+
*
455+
* Then perform the ECC ops.
456+
*
457+
* U = add(T, U)
458+
* T = dbl(T)
459+
*
460+
* Which leaves you with state
461+
*
462+
* k[i] == 0: U = add(R, S), T = dbl(R)
463+
* k[i] == 1: U = add(S, R), T = dbl(S)
464+
*
465+
* Swapping T, U conditionally on k[i] leaves you with state
466+
*
467+
* k[i] == 0: R, S = T, U
468+
* k[i] == 1: R, S = U, T
469+
*
470+
* Which leaves you with state
471+
*
472+
* k[i] == 0: S = add(R, S), R = dbl(R)
473+
* k[i] == 1: R = add(S, R), S = dbl(S)
474+
*
475+
* So we get the same logic, but instead of a branch it's a
476+
* conditional swap, followed by ECC ops, then another conditional swap.
477+
*
478+
* Optimization: The end of iteration i and start of i-1 looks like
479+
*
480+
* ...
481+
* CSWAP(k[i], R, S)
482+
* ECC
483+
* CSWAP(k[i], R, S)
484+
* (next iteration)
485+
* CSWAP(k[i-1], R, S)
486+
* ECC
487+
* CSWAP(k[i-1], R, S)
488+
* ...
489+
*
490+
* So instead of two contiguous swaps, you can merge the condition
491+
* bits and do a single swap.
492+
*
493+
* k[i] k[i-1] Outcome
494+
* 0 0 No Swap
495+
* 0 1 Swap
496+
* 1 0 Swap
497+
* 1 1 No Swap
498+
*
499+
* This is XOR. pbit tracks the previous bit of k.
500+
*/
501+
502+
for (i = cardinality_bits - 1; i >= 0; i--) {
503+
kbit = BN_is_bit_set(k, i) ^ pbit;
504+
EC_POINT_CSWAP(kbit, r, s, group_top, Z_is_one);
505+
if (!EC_POINT_add(group, s, r, s, ctx))
506+
goto err;
507+
if (!EC_POINT_dbl(group, r, r, ctx))
508+
goto err;
509+
/*
510+
* pbit logic merges this cswap with that of the
511+
* next iteration
512+
*/
513+
pbit ^= kbit;
514+
}
515+
/* one final cswap to move the right value into r */
516+
EC_POINT_CSWAP(pbit, r, s, group_top, Z_is_one);
517+
#undef EC_POINT_CSWAP
518+
519+
ret = 1;
520+
521+
err:
522+
EC_POINT_free(s);
523+
BN_CTX_end(ctx);
524+
BN_CTX_free(new_ctx);
525+
526+
return ret;
527+
}
528+
529+
#undef EC_POINT_BN_set_flags
530+
313531
/*
314532
* TODO: table should be optimised for the wNAF-based implementation,
315533
* sometimes smaller windows will give better performance (thus the
@@ -369,6 +587,34 @@ int ec_wNAF_mul(const EC_GROUP *group, EC_POINT *r, const BIGNUM *scalar,
369587
return EC_POINT_set_to_infinity(group, r);
370588
}
371589

590+
if (!BN_is_zero(&group->order) && !BN_is_zero(&group->cofactor)) {
591+
/*-
592+
* Handle the common cases where the scalar is secret, enforcing a constant
593+
* time scalar multiplication algorithm.
594+
*/
595+
if ((scalar != NULL) && (num == 0)) {
596+
/*-
597+
* In this case we want to compute scalar * GeneratorPoint: this
598+
* codepath is reached most prominently by (ephemeral) key generation
599+
* of EC cryptosystems (i.e. ECDSA keygen and sign setup, ECDH
600+
* keygen/first half), where the scalar is always secret. This is why
601+
* we ignore if BN_FLG_CONSTTIME is actually set and we always call the
602+
* constant time version.
603+
*/
604+
return ec_mul_consttime(group, r, scalar, NULL, ctx);
605+
}
606+
if ((scalar == NULL) && (num == 1)) {
607+
/*-
608+
* In this case we want to compute scalar * GenericPoint: this codepath
609+
* is reached most prominently by the second half of ECDH, where the
610+
* secret scalar is multiplied by the peer's public point. To protect
611+
* the secret scalar, we ignore if BN_FLG_CONSTTIME is actually set and
612+
* we always call the constant time version.
613+
*/
614+
return ec_mul_consttime(group, r, scalars[0], points[0], ctx);
615+
}
616+
}
617+
372618
for (i = 0; i < num; i++) {
373619
if (group->meth != points[i]->meth) {
374620
ECerr(EC_F_EC_WNAF_MUL, EC_R_INCOMPATIBLE_OBJECTS);

0 commit comments

Comments
 (0)