… | |
… | |
3 | TODO: bridge mode, finally? |
3 | TODO: bridge mode, finally? |
4 | TODO: gcm mode? |
4 | TODO: gcm mode? |
5 | TODO: replace ripemd160 as the only authentication hash. |
5 | TODO: replace ripemd160 as the only authentication hash. |
6 | TODO: increase rsa size. |
6 | TODO: increase rsa size. |
7 | TODO: replace transport bits by transport endpoint structs? |
7 | TODO: replace transport bits by transport endpoint structs? |
8 | TODO: ecdh to avoid session replay attacks |
|
|
9 | TODO: http://incog-izick.blogspot.de/2011/08/using-openssl-aes-gcm.html |
8 | TODO: http://incog-izick.blogspot.de/2011/08/using-openssl-aes-gcm.html |
10 | TODO: http://stackoverflow.com/questions/12153009/openssl-c-example-of-aes-gcm-using-evp-interfaces |
9 | TODO: http://stackoverflow.com/questions/12153009/openssl-c-example-of-aes-gcm-using-evp-interfaces |
11 | |
10 | |
12 | proposed: 3 types, req, resreq, res |
|
|
13 | |
|
|
14 | req (hmac1) rsa(seqno1 hmac1 aes1 seqno2 hmac2 aes2 auth) ecdh1 |
|
|
15 | res (hmac1) hash(rsa-contents) ecdh2 |
|
|
16 | |
|
|
17 | req hmac0(*) rsa(seqno hmac0 hmac aes auth) hkdf-salt ecdh1 |
|
|
18 | res hmac0(rsa-contents ecdh2) |
|
|
19 | hmac_key = hkdf(hkdf-salt, hmac | ecdh) |
|
|
20 | aes_key = hkdf(hkdf-salt, aes | ecdh) |
|
|
21 | |
|
|
22 | TODO: protocol magic fixen(!!!) |
|
|
23 | TODO: "global" |
|
|
24 | TODO: verify |
11 | TODO: verify |
25 | TODO: make sense of overhead calculation |
12 | TODO: make sense of overhead calculation |
26 | TODO: if-up &c should not be scripts? |
13 | TODO: if-up &c should not be scripts? |
|
|
14 | TODO: ipv6 |
|
|
15 | TODO: gvpectrl should not use default privatekey,. or maybe document it better |
|
|
16 | TODO: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=828336 |
27 | - INCOMPATIBLE CHANGE: core protocol version 1.0. |
17 | - INCOMPATIBLE CHANGE: core protocol version 1.0. |
28 | - INCOMPATIBLE CHANGE: node sextions are now introduced |
18 | - INCOMPATIBLE CHANGE: node sections are now introduced |
29 | with "node nodename", not "node = nodename". |
19 | with "node nodename", not "node = nodename". |
|
|
20 | - add workaround for temporary/rare ENOBUFS condition. |
30 | - while individual packets couldn't be replayed, a whole session |
21 | - while individual packets couldn't be replayed, a whole session |
31 | could be replayed - this has been fixed by an extra key exchange. |
22 | could be replayed - this has been fixed by an extra key exchange. |
|
|
23 | - fix a delete vs. delete [] mismatch in the central logging function. |
32 | - in addition to rsa key exchange and authentication, the handshake now |
24 | - in addition to rsa key exchange and authentication, the handshake now |
33 | adds a diffie-hellman key exchange (using curve25119) for perfect |
25 | adds a diffie-hellman key exchange (using curve25119) for perfect |
34 | forward secrecy. mac and cipher keys are derived using HKDF. |
26 | forward secrecy. mac and cipher keys are derived using HKDF. |
35 | - rsa key sizes are now configurable and larger (default is 3072). |
27 | - rsa key sizes are now configurable and larger (default is 3072). |
36 | correspondingly, the minimum mtu is no longer 296 but 576. |
28 | correspondingly, the minimum mtu is no longer 296 but 576. |
37 | - fixed a potential (unverified) buffer overrun on rsa decryption. |
29 | - fixed a potential (unverified) buffer overrun on rsa decryption. |
|
|
30 | - new per-node low-power setting that tries to reduce cpu/network usage. |
|
|
31 | - router reconnects could cause excessive rekeying on other connections. |
38 | - gvpectrl no longer generates all missing public keys, but |
32 | - gvpectrl no longer generates all missing public keys, but |
39 | only missing private keys. private keys are also put |
33 | only missing private keys. private keys are also put |
40 | into the configured location. |
34 | into the configured location. |
41 | - the pid-file now accepts %s as nodename as elsewhere. |
35 | - the pid-file now accepts %s as nodename as elsewhere. |
|
|
36 | - switch to counter mode (only aes supported at the moment in |
|
|
37 | openssl). this gets rid of the need to generate a random iv, |
|
|
38 | is likely more secure (and, as a side effect, gets rid of |
|
|
39 | slow randomness generation. counter mode is often faster |
|
|
40 | then cbc mode as well, and packets are smaller). |
42 | - no longer use RAND_bytes to generate session keys - you NEED |
41 | - no longer use RAND_bytes to generate session keys - you NEED |
43 | a real source of entropy now (e.g. egd or /dev/random - see the |
42 | a real source of entropy now (e.g. egd or /dev/random - see the |
44 | openssl documentation). |
43 | openssl documentation). |
45 | - multiple node statements for the same node are now supported |
44 | - multiple node statements for the same node are now supported |
46 | and will be merged. |
45 | and will be merged. |
47 | - a new directive "global" switches back to the global section |
46 | - a new directive "global" switches back to the global section |
48 | of the config file. |
47 | of the config file. |
49 | - 12 random prefix bytes are now properly supported, leading to |
|
|
50 | a fully random IV. |
|
|
51 | - use aes with a random key in counter-mode to generate IVs, |
|
|
52 | for speed reasons (generating 12 random bytes with openssl can |
|
|
53 | take longer than encrypting and mac'ing a 1.5kb packet, and IVs |
|
|
54 | do not need to be cryptographically strong random numbers |
|
|
55 | (and in fact, shouldn't be)). |
|
|
56 | - if-up scripts can now be specified with absolute paths. |
48 | - if-up scripts can now be specified with absolute paths. |
57 | - new global option: serial, to detect configuration mismatches. |
49 | - new global option: serial, to detect configuration mismatches. |
58 | - use HKDF as authentication proof, not HMAC or a plain hash |
50 | - use HKDF as authentication proof, not HMAC or a plain hash |
59 | (hint by Ilmari Karonen). |
51 | (hint by Ilmari Karonen). |
|
|
52 | - during rekeying or conenction establishments, hmac authentication |
|
|
53 | errors could occur and reset the connection. Transient hmac |
|
|
54 | authentication errors are now being ignored for 3 seconds. |
|
|
55 | - log the reason for a conneciton loss. |
|
|
56 | - use a (hopefully) constant time memcmp to compare internal secrets. |
|
|
57 | - fix a (harmless) errornous out of bounds stack read that would trigger |
|
|
58 | gcc's -fsanitize=address. |
|
|
59 | - bump old packet window size from 512 to 65536. |
|
|
60 | - update for big changes in openssl 1.1 API, wrap primitives |
|
|
61 | to make further changes easier. |
|
|
62 | - correctly check return values for openssl 1.0.0 and later. |
60 | |
63 | |
61 | 2.25 Sat Jul 13 06:42:33 CEST 2013 |
64 | 2.25 Sat Jul 13 06:42:33 CEST 2013 |
62 | - INCOMPATIBLE CHANGE: no longer enable udp protocol if no other |
65 | - INCOMPATIBLE CHANGE: no longer enable udp protocol if no other |
63 | protocols are enabled - this is necessary when you have nodes with |
66 | protocols are enabled - this is necessary when you have nodes with |
64 | completely unknown protocols, to force mediated connection requests. |
67 | completely unknown protocols, to force mediated connection requests. |