Mercurial > dropbear
view libtomcrypt/src/pk/asn1/der/object_identifier/der_decode_object_identifier.c @ 1790:42745af83b7d
Introduce extra delay before closing unauthenticated sessions
To make it harder for attackers, introduce a delay to keep an
unauthenticated session open a bit longer, thus blocking a connection
slot until after the delay.
Without this, while there is a limit on the amount of attempts an attacker
can make at the same time (MAX_UNAUTH_PER_IP), the time taken by dropbear to
handle one attempt is still short and thus for each of the allowed parallel
attempts many attempts can be chained one after the other. The attempt rate
is then:
"MAX_UNAUTH_PER_IP / <process time of one attempt>".
With the delay, this rate becomes:
"MAX_UNAUTH_PER_IP / UNAUTH_CLOSE_DELAY".
author | Thomas De Schampheleire <thomas.de_schampheleire@nokia.com> |
---|---|
date | Wed, 15 Feb 2017 13:53:04 +0100 |
parents | 6dba84798cd5 |
children |
line wrap: on
line source
/* LibTomCrypt, modular cryptographic library -- Tom St Denis * * LibTomCrypt is a library that provides various cryptographic * algorithms in a highly modular and flexible manner. * * The library is free for all purposes without any express * guarantee it works. */ #include "tomcrypt.h" /** @file der_decode_object_identifier.c ASN.1 DER, Decode Object Identifier, Tom St Denis */ #ifdef LTC_DER /** Decode OID data and store the array of integers in words @param in The OID DER encoded data @param inlen The length of the OID data @param words [out] The destination of the OID words @param outlen [in/out] The number of OID words @return CRYPT_OK if successful */ int der_decode_object_identifier(const unsigned char *in, unsigned long inlen, unsigned long *words, unsigned long *outlen) { unsigned long x, y, t, len; int err; LTC_ARGCHK(in != NULL); LTC_ARGCHK(words != NULL); LTC_ARGCHK(outlen != NULL); /* header is at least 3 bytes */ if (inlen < 3) { return CRYPT_INVALID_PACKET; } /* must be room for at least two words */ if (*outlen < 2) { *outlen = 2; return CRYPT_BUFFER_OVERFLOW; } /* decode the packet header */ x = 0; if ((in[x++] & 0x1F) != 0x06) { return CRYPT_INVALID_PACKET; } /* get the length */ if (in[x] < 128) { len = in[x++]; } else { if (in[x] < 0x81 || in[x] > 0x82) { return CRYPT_INVALID_PACKET; } y = in[x++] & 0x7F; len = 0; while (y--) { len = (len << 8) | (unsigned long)in[x++]; } } if (len < 1 || (len + x) > inlen) { return CRYPT_INVALID_PACKET; } /* decode words */ y = 0; t = 0; while (len--) { t = (t << 7) | (in[x] & 0x7F); if (!(in[x++] & 0x80)) { /* store t */ if (y >= *outlen) { y++; } else { if (y == 0) { words[0] = t / 40; words[1] = t % 40; y = 2; } else { words[y++] = t; } } t = 0; } } if (y > *outlen) { err = CRYPT_BUFFER_OVERFLOW; } else { err = CRYPT_OK; } *outlen = y; return err; } #endif /* ref: $Format:%D$ */ /* git commit: $Format:%H$ */ /* commit time: $Format:%ai$ */