Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
/*
|
|
|
|
sptps.c -- Simple Peer-to-Peer Security
|
|
|
|
Copyright (C) 2011 Guus Sliepen <guus@tinc-vpn.org>,
|
|
|
|
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
|
|
it under the terms of the GNU General Public License as published by
|
|
|
|
the Free Software Foundation; either version 2 of the License, or
|
|
|
|
(at your option) any later version.
|
|
|
|
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
GNU General Public License for more details.
|
|
|
|
|
|
|
|
You should have received a copy of the GNU General Public License along
|
|
|
|
with this program; if not, write to the Free Software Foundation, Inc.,
|
|
|
|
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "system.h"
|
|
|
|
|
|
|
|
#include "cipher.h"
|
|
|
|
#include "crypto.h"
|
|
|
|
#include "digest.h"
|
|
|
|
#include "ecdh.h"
|
|
|
|
#include "ecdsa.h"
|
|
|
|
#include "prf.h"
|
|
|
|
#include "sptps.h"
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
/*
|
|
|
|
Nonce MUST be exchanged first (done)
|
|
|
|
Signatures MUST be done over both nonces, to guarantee the signature is fresh
|
|
|
|
Otherwise: if ECDHE key of one side is compromised, it can be reused!
|
|
|
|
|
|
|
|
Add explicit tag to beginning of structure to distinguish the client and server when signing. (done)
|
|
|
|
|
|
|
|
Sign all handshake messages up to ECDHE kex with long-term public keys. (done)
|
|
|
|
|
|
|
|
HMACed KEX finished message to prevent downgrade attacks and prove you have the right key material (done by virtue of ECDSA over the whole ECDHE exchange?)
|
|
|
|
|
|
|
|
Explicit close message needs to be added.
|
|
|
|
|
|
|
|
Maybe do add some alert messages to give helpful error messages? Not more than TLS sends.
|
|
|
|
|
2011-10-06 07:34:34 +00:00
|
|
|
Use counter mode instead of OFB. (done)
|
2011-10-05 20:00:51 +00:00
|
|
|
|
|
|
|
Make sure ECC operations are fixed time (aka prevent side-channel attacks).
|
|
|
|
*/
|
|
|
|
|
|
|
|
// Log an error message.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
static bool error(sptps_t *s, int s_errno, const char *msg) {
|
|
|
|
fprintf(stderr, "SPTPS error: %s\n", msg);
|
|
|
|
errno = s_errno;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2012-03-18 15:42:02 +00:00
|
|
|
// Send a record (datagram version, accepts all record types, handles encryption and authentication).
|
|
|
|
static bool send_record_priv_datagram(sptps_t *s, uint8_t type, const char *data, uint16_t len) {
|
|
|
|
char buffer[len + 23UL];
|
|
|
|
|
|
|
|
// Create header with sequence number, length and record type
|
|
|
|
uint32_t seqno = htonl(s->outseqno++);
|
|
|
|
uint16_t netlen = htons(len);
|
|
|
|
|
|
|
|
memcpy(buffer, &netlen, 2);
|
|
|
|
memcpy(buffer + 2, &seqno, 4);
|
|
|
|
buffer[6] = type;
|
|
|
|
|
|
|
|
// Add plaintext (TODO: avoid unnecessary copy)
|
|
|
|
memcpy(buffer + 7, data, len);
|
|
|
|
|
|
|
|
if(s->outstate) {
|
|
|
|
// If first handshake has finished, encrypt and HMAC
|
|
|
|
cipher_set_counter(&s->outcipher, &seqno, sizeof seqno);
|
|
|
|
if(!cipher_counter_xor(&s->outcipher, buffer + 6, len + 1UL, buffer + 6))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if(!digest_create(&s->outdigest, buffer, len + 7UL, buffer + 7UL + len))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return s->send_data(s->handle, buffer + 2, len + 21UL);
|
|
|
|
} else {
|
|
|
|
// Otherwise send as plaintext
|
|
|
|
return s->send_data(s->handle, buffer + 2, len + 5UL);
|
|
|
|
}
|
|
|
|
}
|
2011-10-05 20:00:51 +00:00
|
|
|
// Send a record (private version, accepts all record types, handles encryption and authentication).
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
static bool send_record_priv(sptps_t *s, uint8_t type, const char *data, uint16_t len) {
|
2012-03-18 15:42:02 +00:00
|
|
|
if(s->datagram)
|
|
|
|
return send_record_priv_datagram(s, type, data, len);
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
char buffer[len + 23UL];
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
|
|
|
// Create header with sequence number, length and record type
|
|
|
|
uint32_t seqno = htonl(s->outseqno++);
|
|
|
|
uint16_t netlen = htons(len);
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
memcpy(buffer, &seqno, 4);
|
|
|
|
memcpy(buffer + 4, &netlen, 2);
|
|
|
|
buffer[6] = type;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
|
|
|
// Add plaintext (TODO: avoid unnecessary copy)
|
2012-02-25 14:18:15 +00:00
|
|
|
memcpy(buffer + 7, data, len);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
if(s->outstate) {
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// If first handshake has finished, encrypt and HMAC
|
2012-02-25 14:18:15 +00:00
|
|
|
if(!cipher_counter_xor(&s->outcipher, buffer + 4, len + 3UL, buffer + 4))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
if(!digest_create(&s->outdigest, buffer, len + 7UL, buffer + 7UL + len))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
return s->send_data(s->handle, buffer + 4, len + 19UL);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
} else {
|
|
|
|
// Otherwise send as plaintext
|
2012-02-25 14:18:15 +00:00
|
|
|
return s->send_data(s->handle, buffer + 4, len + 3UL);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Send an application record.
|
2012-02-26 11:33:16 +00:00
|
|
|
bool sptps_send_record(sptps_t *s, uint8_t type, const char *data, uint16_t len) {
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// Sanity checks: application cannot send data before handshake is finished,
|
|
|
|
// and only record types 0..127 are allowed.
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!s->outstate)
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return error(s, EINVAL, "Handshake phase not finished yet");
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
if(type >= SPTPS_HANDSHAKE)
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return error(s, EINVAL, "Invalid application record type");
|
|
|
|
|
|
|
|
return send_record_priv(s, type, data, len);
|
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Send a Key EXchange record, containing a random nonce and an ECDHE public key.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
static bool send_kex(sptps_t *s) {
|
|
|
|
size_t keylen = ECDH_SIZE;
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Make room for our KEX message, which we will keep around since send_sig() needs it.
|
2012-03-18 15:42:02 +00:00
|
|
|
if(s->mykex)
|
|
|
|
abort();
|
2011-10-05 20:00:51 +00:00
|
|
|
s->mykex = realloc(s->mykex, 1 + 32 + keylen);
|
|
|
|
if(!s->mykex)
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return error(s, errno, strerror(errno));
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Set version byte to zero.
|
|
|
|
s->mykex[0] = SPTPS_VERSION;
|
|
|
|
|
|
|
|
// Create a random nonce.
|
|
|
|
randomize(s->mykex + 1, 32);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
|
|
|
// Create a new ECDH public key.
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!ecdh_generate_public(&s->ecdh, s->mykex + 1 + 32))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
return send_record_priv(s, SPTPS_HANDSHAKE, s->mykex, 1 + 32 + keylen);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Send a SIGnature record, containing an ECDSA signature over both KEX records.
|
|
|
|
static bool send_sig(sptps_t *s) {
|
|
|
|
size_t keylen = ECDH_SIZE;
|
|
|
|
size_t siglen = ecdsa_size(&s->mykey);
|
|
|
|
|
|
|
|
// Concatenate both KEX messages, plus tag indicating if it is from the connection originator
|
|
|
|
char msg[(1 + 32 + keylen) * 2 + 1];
|
|
|
|
char sig[siglen];
|
|
|
|
|
|
|
|
msg[0] = s->initiator;
|
|
|
|
memcpy(msg + 1, s->mykex, 1 + 32 + keylen);
|
|
|
|
memcpy(msg + 2 + 32 + keylen, s->hiskex, 1 + 32 + keylen);
|
|
|
|
|
|
|
|
// Sign the result.
|
|
|
|
if(!ecdsa_sign(&s->mykey, msg, sizeof msg, sig))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Send the SIG exchange record.
|
|
|
|
return send_record_priv(s, SPTPS_HANDSHAKE, sig, sizeof sig);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Generate key material from the shared secret created from the ECDHE key exchange.
|
|
|
|
static bool generate_key_material(sptps_t *s, const char *shared, size_t len) {
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// Initialise cipher and digest structures if necessary
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!s->outstate) {
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
bool result
|
2011-10-06 07:34:34 +00:00
|
|
|
= cipher_open_by_name(&s->incipher, "aes-256-ecb")
|
|
|
|
&& cipher_open_by_name(&s->outcipher, "aes-256-ecb")
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
&& digest_open_by_name(&s->indigest, "sha256", 16)
|
|
|
|
&& digest_open_by_name(&s->outdigest, "sha256", 16);
|
|
|
|
if(!result)
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Allocate memory for key material
|
|
|
|
size_t keylen = digest_keylength(&s->indigest) + digest_keylength(&s->outdigest) + cipher_keylength(&s->incipher) + cipher_keylength(&s->outcipher);
|
|
|
|
|
|
|
|
s->key = realloc(s->key, keylen);
|
|
|
|
if(!s->key)
|
|
|
|
return error(s, errno, strerror(errno));
|
|
|
|
|
|
|
|
// Create the HMAC seed, which is "key expansion" + session label + server nonce + client nonce
|
|
|
|
char seed[s->labellen + 64 + 13];
|
|
|
|
strcpy(seed, "key expansion");
|
|
|
|
if(s->initiator) {
|
2011-10-05 20:00:51 +00:00
|
|
|
memcpy(seed + 13, s->mykex + 1, 32);
|
|
|
|
memcpy(seed + 45, s->hiskex + 1, 32);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
} else {
|
2011-10-05 20:00:51 +00:00
|
|
|
memcpy(seed + 13, s->hiskex + 1, 32);
|
|
|
|
memcpy(seed + 45, s->mykex + 1, 32);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
}
|
|
|
|
memcpy(seed + 78, s->label, s->labellen);
|
|
|
|
|
|
|
|
// Use PRF to generate the key material
|
|
|
|
if(!prf(shared, len, seed, s->labellen + 64 + 13, s->key, keylen))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Send an ACKnowledgement record.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
static bool send_ack(sptps_t *s) {
|
2011-10-05 20:00:51 +00:00
|
|
|
return send_record_priv(s, SPTPS_HANDSHAKE, "", 0);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Receive an ACKnowledgement record.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
static bool receive_ack(sptps_t *s, const char *data, uint16_t len) {
|
|
|
|
if(len)
|
2011-10-06 13:32:12 +00:00
|
|
|
return error(s, EIO, "Invalid ACK record length");
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
2011-10-06 13:32:12 +00:00
|
|
|
if(s->initiator) {
|
|
|
|
bool result
|
|
|
|
= cipher_set_counter_key(&s->incipher, s->key)
|
|
|
|
&& digest_set_key(&s->indigest, s->key + cipher_keylength(&s->incipher), digest_keylength(&s->indigest));
|
|
|
|
if(!result)
|
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
bool result
|
|
|
|
= cipher_set_counter_key(&s->incipher, s->key + cipher_keylength(&s->outcipher) + digest_keylength(&s->outdigest))
|
|
|
|
&& digest_set_key(&s->indigest, s->key + cipher_keylength(&s->outcipher) + digest_keylength(&s->outdigest) + cipher_keylength(&s->incipher), digest_keylength(&s->indigest));
|
|
|
|
if(!result)
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
free(s->key);
|
|
|
|
s->key = NULL;
|
|
|
|
s->instate = true;
|
|
|
|
|
|
|
|
return true;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Receive a Key EXchange record, respond by sending a SIG record.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
static bool receive_kex(sptps_t *s, const char *data, uint16_t len) {
|
2011-10-05 20:00:51 +00:00
|
|
|
// Verify length of the HELLO record
|
|
|
|
if(len != 1 + 32 + ECDH_SIZE)
|
|
|
|
return error(s, EIO, "Invalid KEX record length");
|
|
|
|
|
|
|
|
// Ignore version number for now.
|
|
|
|
|
|
|
|
// Make a copy of the KEX message, send_sig() and receive_sig() need it
|
2012-03-18 15:42:02 +00:00
|
|
|
if(s->hiskex)
|
|
|
|
abort();
|
2011-10-05 20:00:51 +00:00
|
|
|
s->hiskex = realloc(s->hiskex, len);
|
|
|
|
if(!s->hiskex)
|
|
|
|
return error(s, errno, strerror(errno));
|
|
|
|
|
|
|
|
memcpy(s->hiskex, data, len);
|
|
|
|
|
|
|
|
return send_sig(s);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Receive a SIGnature record, verify it, if it passed, compute the shared secret and calculate the session keys.
|
|
|
|
static bool receive_sig(sptps_t *s, const char *data, uint16_t len) {
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
size_t keylen = ECDH_SIZE;
|
|
|
|
size_t siglen = ecdsa_size(&s->hiskey);
|
|
|
|
|
|
|
|
// Verify length of KEX record.
|
2011-10-05 20:00:51 +00:00
|
|
|
if(len != siglen)
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return error(s, EIO, "Invalid KEX record length");
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Concatenate both KEX messages, plus tag indicating if it is from the connection originator
|
|
|
|
char msg[(1 + 32 + keylen) * 2 + 1];
|
|
|
|
|
|
|
|
msg[0] = !s->initiator;
|
|
|
|
memcpy(msg + 1, s->hiskex, 1 + 32 + keylen);
|
|
|
|
memcpy(msg + 2 + 32 + keylen, s->mykex, 1 + 32 + keylen);
|
|
|
|
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// Verify signature.
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!ecdsa_verify(&s->hiskey, msg, sizeof msg, data))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
|
|
|
|
// Compute shared secret.
|
|
|
|
char shared[ECDH_SHARED_SIZE];
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!ecdh_compute_shared(&s->ecdh, s->hiskex + 1 + 32, shared))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
|
|
|
|
// Generate key material from shared secret.
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!generate_key_material(s, shared, sizeof shared))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
|
2011-10-06 13:32:12 +00:00
|
|
|
free(s->mykex);
|
|
|
|
free(s->hiskex);
|
|
|
|
|
|
|
|
s->mykex = NULL;
|
|
|
|
s->hiskex = NULL;
|
|
|
|
|
|
|
|
// Send cipher change record
|
2012-03-18 16:46:30 +00:00
|
|
|
if(s->outstate && !send_ack(s))
|
2011-10-06 13:32:12 +00:00
|
|
|
return false;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// TODO: only set new keys after ACK has been set/received
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
if(s->initiator) {
|
|
|
|
bool result
|
2011-10-06 13:32:12 +00:00
|
|
|
= cipher_set_counter_key(&s->outcipher, s->key + cipher_keylength(&s->incipher) + digest_keylength(&s->indigest))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
&& digest_set_key(&s->outdigest, s->key + cipher_keylength(&s->incipher) + digest_keylength(&s->indigest) + cipher_keylength(&s->outcipher), digest_keylength(&s->outdigest));
|
|
|
|
if(!result)
|
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
bool result
|
2011-10-06 07:34:34 +00:00
|
|
|
= cipher_set_counter_key(&s->outcipher, s->key)
|
2011-10-06 13:32:12 +00:00
|
|
|
&& digest_set_key(&s->outdigest, s->key + cipher_keylength(&s->outcipher), digest_keylength(&s->outdigest));
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
if(!result)
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Force another Key EXchange (for testing purposes).
|
2012-02-26 11:33:16 +00:00
|
|
|
bool sptps_force_kex(sptps_t *s) {
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!s->outstate || s->state != SPTPS_SECONDARY_KEX)
|
|
|
|
return error(s, EINVAL, "Cannot force KEX in current state");
|
|
|
|
|
|
|
|
s->state = SPTPS_KEX;
|
|
|
|
return send_kex(s);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Receive a handshake record.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
static bool receive_handshake(sptps_t *s, const char *data, uint16_t len) {
|
|
|
|
// Only a few states to deal with handshaking.
|
2011-10-05 20:00:51 +00:00
|
|
|
fprintf(stderr, "Received handshake message, current state %d\n", s->state);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
switch(s->state) {
|
2011-10-05 20:00:51 +00:00
|
|
|
case SPTPS_SECONDARY_KEX:
|
|
|
|
// We receive a secondary KEX request, first respond by sending our own.
|
|
|
|
if(!send_kex(s))
|
|
|
|
return false;
|
|
|
|
case SPTPS_KEX:
|
|
|
|
// We have sent our KEX request, we expect our peer to sent one as well.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
if(!receive_kex(s, data, len))
|
|
|
|
return false;
|
2011-10-05 20:00:51 +00:00
|
|
|
s->state = SPTPS_SIG;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return true;
|
2011-10-05 20:00:51 +00:00
|
|
|
case SPTPS_SIG:
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// If we already sent our secondary public ECDH key, we expect the peer to send his.
|
2011-10-05 20:00:51 +00:00
|
|
|
if(!receive_sig(s, data, len))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
2012-03-18 16:46:30 +00:00
|
|
|
if(s->outstate)
|
|
|
|
s->state = SPTPS_ACK;
|
|
|
|
else {
|
|
|
|
s->outstate = true;
|
|
|
|
if(!receive_ack(s, NULL, 0))
|
|
|
|
return false;
|
|
|
|
s->receive_record(s->handle, SPTPS_HANDSHAKE, NULL, 0);
|
|
|
|
s->state = SPTPS_SECONDARY_KEX;
|
|
|
|
}
|
|
|
|
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return true;
|
2011-10-05 20:00:51 +00:00
|
|
|
case SPTPS_ACK:
|
|
|
|
// We expect a handshake message to indicate transition to the new keys.
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
if(!receive_ack(s, data, len))
|
|
|
|
return false;
|
2012-02-25 17:25:21 +00:00
|
|
|
s->receive_record(s->handle, SPTPS_HANDSHAKE, NULL, 0);
|
2011-10-05 20:00:51 +00:00
|
|
|
s->state = SPTPS_SECONDARY_KEX;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return true;
|
2011-10-05 20:00:51 +00:00
|
|
|
// TODO: split ACK into a VERify and ACK?
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
default:
|
|
|
|
return error(s, EIO, "Invalid session state");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-03-18 15:42:02 +00:00
|
|
|
// Receive incoming data, datagram version.
|
|
|
|
static bool sptps_receive_data_datagram(sptps_t *s, const char *data, size_t len) {
|
|
|
|
if(len < (s->instate ? 21 : 5))
|
|
|
|
return error(s, EIO, "Received short packet");
|
|
|
|
|
|
|
|
uint32_t seqno;
|
|
|
|
memcpy(&seqno, data, 4);
|
|
|
|
seqno = ntohl(seqno);
|
|
|
|
|
|
|
|
if(!s->instate) {
|
|
|
|
if(seqno != s->inseqno) {
|
|
|
|
fprintf(stderr, "Received invalid packet seqno: %d != %d\n", seqno, s->inseqno);
|
|
|
|
return error(s, EIO, "Invalid packet seqno");
|
|
|
|
}
|
|
|
|
|
|
|
|
s->inseqno = seqno + 1;
|
|
|
|
|
|
|
|
uint8_t type = data[4];
|
|
|
|
|
|
|
|
if(type != SPTPS_HANDSHAKE)
|
|
|
|
return error(s, EIO, "Application record received before handshake finished");
|
|
|
|
|
|
|
|
return receive_handshake(s, data + 5, len - 5);
|
|
|
|
}
|
|
|
|
|
|
|
|
if(seqno < s->inseqno) {
|
|
|
|
fprintf(stderr, "Received late or replayed packet: %d < %d\n", seqno, s->inseqno);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if(seqno > s->inseqno)
|
|
|
|
fprintf(stderr, "Missed %d packets\n", seqno - s->inseqno);
|
|
|
|
|
|
|
|
s->inseqno = seqno + 1;
|
|
|
|
|
|
|
|
uint16_t netlen = htons(len - 21);
|
|
|
|
|
|
|
|
char buffer[len + 23];
|
|
|
|
|
|
|
|
memcpy(buffer, &netlen, 2);
|
|
|
|
memcpy(buffer + 2, data, len);
|
|
|
|
|
|
|
|
memcpy(&seqno, buffer + 2, 4);
|
|
|
|
|
|
|
|
// Check HMAC and decrypt.
|
|
|
|
if(!digest_verify(&s->indigest, buffer, len - 14, buffer + len - 14))
|
|
|
|
return error(s, EIO, "Invalid HMAC");
|
|
|
|
|
|
|
|
cipher_set_counter(&s->incipher, &seqno, sizeof seqno);
|
|
|
|
if(!cipher_counter_xor(&s->incipher, buffer + 6, len - 4, buffer + 6))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
// Append a NULL byte for safety.
|
|
|
|
buffer[len - 14] = 0;
|
|
|
|
|
|
|
|
uint8_t type = buffer[6];
|
|
|
|
|
|
|
|
if(type < SPTPS_HANDSHAKE) {
|
|
|
|
if(!s->instate)
|
|
|
|
return error(s, EIO, "Application record received before handshake finished");
|
|
|
|
if(!s->receive_record(s->handle, type, buffer + 7, len - 21))
|
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
return error(s, EIO, "Invalid record type");
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
2011-10-05 20:00:51 +00:00
|
|
|
// Receive incoming data. Check if it contains a complete record, if so, handle it.
|
2012-02-26 11:33:16 +00:00
|
|
|
bool sptps_receive_data(sptps_t *s, const char *data, size_t len) {
|
2012-03-18 15:42:02 +00:00
|
|
|
if(s->datagram)
|
|
|
|
return sptps_receive_data_datagram(s, data, len);
|
|
|
|
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
while(len) {
|
|
|
|
// First read the 2 length bytes.
|
|
|
|
if(s->buflen < 6) {
|
|
|
|
size_t toread = 6 - s->buflen;
|
|
|
|
if(toread > len)
|
|
|
|
toread = len;
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
memcpy(s->inbuf + s->buflen, data, toread);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
|
|
|
s->buflen += toread;
|
|
|
|
len -= toread;
|
|
|
|
data += toread;
|
2012-02-25 14:18:15 +00:00
|
|
|
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// Exit early if we don't have the full length.
|
|
|
|
if(s->buflen < 6)
|
|
|
|
return true;
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
// Decrypt the length bytes
|
|
|
|
|
|
|
|
if(s->instate) {
|
|
|
|
if(!cipher_counter_xor(&s->incipher, s->inbuf + 4, 2, &s->reclen))
|
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
memcpy(&s->reclen, s->inbuf + 4, 2);
|
|
|
|
}
|
|
|
|
|
|
|
|
s->reclen = ntohs(s->reclen);
|
|
|
|
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// If we have the length bytes, ensure our buffer can hold the whole request.
|
2012-02-25 14:18:15 +00:00
|
|
|
s->inbuf = realloc(s->inbuf, s->reclen + 23UL);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
if(!s->inbuf)
|
|
|
|
return error(s, errno, strerror(errno));
|
|
|
|
|
|
|
|
// Add sequence number.
|
|
|
|
uint32_t seqno = htonl(s->inseqno++);
|
|
|
|
memcpy(s->inbuf, &seqno, 4);
|
|
|
|
|
|
|
|
// Exit early if we have no more data to process.
|
|
|
|
if(!len)
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Read up to the end of the record.
|
2012-02-25 14:18:15 +00:00
|
|
|
size_t toread = s->reclen + (s->instate ? 23UL : 7UL) - s->buflen;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
if(toread > len)
|
|
|
|
toread = len;
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
memcpy(s->inbuf + s->buflen, data, toread);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
s->buflen += toread;
|
|
|
|
len -= toread;
|
|
|
|
data += toread;
|
|
|
|
|
|
|
|
// If we don't have a whole record, exit.
|
2012-02-25 14:18:15 +00:00
|
|
|
if(s->buflen < s->reclen + (s->instate ? 23UL : 7UL))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return true;
|
|
|
|
|
2012-02-25 14:18:15 +00:00
|
|
|
// Check HMAC and decrypt.
|
|
|
|
if(s->instate) {
|
|
|
|
if(!digest_verify(&s->indigest, s->inbuf, s->reclen + 7UL, s->inbuf + s->reclen + 7UL))
|
|
|
|
return error(s, EIO, "Invalid HMAC");
|
|
|
|
|
|
|
|
if(!cipher_counter_xor(&s->incipher, s->inbuf + 6UL, s->reclen + 1UL, s->inbuf + 6UL))
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Append a NULL byte for safety.
|
|
|
|
s->inbuf[s->reclen + 7UL] = 0;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
|
|
|
uint8_t type = s->inbuf[6];
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
if(type < SPTPS_HANDSHAKE) {
|
2011-10-06 13:32:12 +00:00
|
|
|
if(!s->instate)
|
|
|
|
return error(s, EIO, "Application record received before handshake finished");
|
2012-02-25 14:18:15 +00:00
|
|
|
if(!s->receive_record(s->handle, type, s->inbuf + 7, s->reclen))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
2011-10-05 20:00:51 +00:00
|
|
|
} else if(type == SPTPS_HANDSHAKE) {
|
2012-02-25 14:18:15 +00:00
|
|
|
if(!receive_handshake(s, s->inbuf + 7, s->reclen))
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
return error(s, EIO, "Invalid record type");
|
|
|
|
}
|
|
|
|
|
|
|
|
s->buflen = 4;
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Start a SPTPS session.
|
2012-03-18 15:42:02 +00:00
|
|
|
bool sptps_start(sptps_t *s, void *handle, bool initiator, bool datagram, ecdsa_t mykey, ecdsa_t hiskey, const char *label, size_t labellen, send_data_t send_data, receive_record_t receive_record) {
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// Initialise struct sptps
|
|
|
|
memset(s, 0, sizeof *s);
|
|
|
|
|
|
|
|
s->handle = handle;
|
|
|
|
s->initiator = initiator;
|
2012-03-18 15:42:02 +00:00
|
|
|
s->datagram = datagram;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
s->mykey = mykey;
|
|
|
|
s->hiskey = hiskey;
|
|
|
|
|
|
|
|
s->label = malloc(labellen);
|
|
|
|
if(!s->label)
|
|
|
|
return error(s, errno, strerror(errno));
|
|
|
|
|
2012-03-18 15:42:02 +00:00
|
|
|
if(!datagram) {
|
|
|
|
s->inbuf = malloc(7);
|
|
|
|
if(!s->inbuf)
|
|
|
|
return error(s, errno, strerror(errno));
|
|
|
|
s->buflen = 4;
|
|
|
|
memset(s->inbuf, 0, 4);
|
|
|
|
}
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
|
|
|
|
memcpy(s->label, label, labellen);
|
|
|
|
s->labellen = labellen;
|
|
|
|
|
|
|
|
s->send_data = send_data;
|
|
|
|
s->receive_record = receive_record;
|
|
|
|
|
|
|
|
// Do first KEX immediately
|
2011-10-05 20:00:51 +00:00
|
|
|
s->state = SPTPS_KEX;
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
return send_kex(s);
|
|
|
|
}
|
|
|
|
|
2011-10-05 20:00:51 +00:00
|
|
|
// Stop a SPTPS session.
|
2012-02-26 11:33:16 +00:00
|
|
|
bool sptps_stop(sptps_t *s) {
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
// Clean up any resources.
|
|
|
|
ecdh_free(&s->ecdh);
|
|
|
|
free(s->inbuf);
|
2011-10-05 20:00:51 +00:00
|
|
|
free(s->mykex);
|
|
|
|
free(s->hiskex);
|
Start of "Simple Peer-To-Peer Security" protocol.
Encryption and authentication of the meta connection is spread out over
meta.c and protocol_auth.c. The new protocol was added there as well,
leading to spaghetti code. To improve things, the new protocol will now
be implemented in sptps.[ch].
The goal is to have a very simplified version of TLS. There is a record
layer, and there are only two record types: application data and
handshake messages. The handshake message contains a random nonce, an
ephemeral ECDH public key, and an ECDSA signature over the former. After
the ECDH public keys are exchanged, a shared secret is calculated, and a
TLS style PRF is used to generate the key material for the cipher and
HMAC algorithm, and further communication is encrypted and authenticated.
A lot of the simplicity comes from the fact that both sides must have
each other's public keys in advance, and there are no options to choose.
There will be one fixed cipher suite, and both peers always authenticate
each other. (Inspiration taken from Ian Grigg's hypotheses[0].)
There might be some compromise in the future, to enable or disable
encryption, authentication and compression, but there will be no choice
of algorithms. This will allow SPTPS to be built with a few embedded
crypto algorithms instead of linking with huge crypto libraries.
The API is also kept simple. There is a start and a stop function. All
data necessary to make the connection work is passed in the start
function. Instead having both send- and receive-record functions, there
is a send-record function and a receive-data function. The latter will
pass protocol data received from the peer to the SPTPS implementation,
which will in turn call a receive-record callback function when
necessary. This hides all the handshaking from the application, and is
completely independent from any event loop or socket characteristics.
[0] http://iang.org/ssl/hn_hypotheses_in_secure_protocol_design.html
2011-07-24 13:44:51 +00:00
|
|
|
free(s->key);
|
|
|
|
free(s->label);
|
|
|
|
return true;
|
|
|
|
}
|