diff options
Diffstat (limited to 'sm/certcheck.c')
| -rw-r--r-- | sm/certcheck.c | 40 |
1 files changed, 39 insertions, 1 deletions
diff --git a/sm/certcheck.c b/sm/certcheck.c index 68185332e..14f78dbe6 100644 --- a/sm/certcheck.c +++ b/sm/certcheck.c @@ -300,7 +300,45 @@ extract_pss_params (gcry_sexp_t s_sig, int *r_algo, unsigned int *r_saltlen) { return gpg_error (GPG_ERR_DIGEST_ALGO); } - /* log_debug ("PSS hash=%d saltlen=%u\n", *r_algo, *r_saltlen); */ + + /* PSS has no hash function firewall like PKCS#1 and thus offers + * a path for hash algorithm replacement. To avoid this it makes + * sense to restrict the allowed hash algorithms and also allow only + * matching salt lengths. According to Peter Gutmann: + * "Beware of bugs in the above signature scheme; + * I have only proved it secure, not implemented it" + * - Apologies to Donald Knuth. + * https://www.metzdowd.com/pipermail/cryptography/2019-November/035449.html + * + * Given the set of supported algorithms currently available in + * Libgcrypt and the extra hash checks we have in some compliance + * modes, it would be hard to trick gpgsm to verify a forged + * signature. However, if eventually someone adds the xor256 hash + * algorithm (1.3.6.1.4.1.3029.3.2) to Libgcrypt we would be doomed. + */ + switch (*r_algo) + { + case GCRY_MD_SHA1: + case GCRY_MD_SHA256: + case GCRY_MD_SHA384: + case GCRY_MD_SHA512: + case GCRY_MD_SHA3_256: + case GCRY_MD_SHA3_384: + case GCRY_MD_SHA3_512: + break; + default: + log_error ("PSS hash algorithm '%s' rejected\n", + gcry_md_algo_name (*r_algo)); + return gpg_error (GPG_ERR_DIGEST_ALGO); + } + + if (gcry_md_get_algo_dlen (*r_algo) != *r_saltlen) + { + log_error ("PSS hash algorithm '%s' rejected due to salt length %u\n", + gcry_md_algo_name (*r_algo), *r_saltlen); + return gpg_error (GPG_ERR_DIGEST_ALGO); + } + return 0; } |