Disable TLSv1.1 and TLSv1.2 in the Firefox helper

With #11253 (closed), Tor Browser's Firefox config has TLSv1.1 and TLSv1.2 turned on. If meek-http-helper (browser TLS camouflage) sends Firefox 24 ciphersuites but uses TLSv1.1 or TLSv1.2, then it will look weird, because as I understand it, mainline Firefox 24 has TLSv1.1 and TLSv1.2 disabled. ([[doc/meek#Sampleclienthellos]] corroborates that ordinary Firefox 24 uses TLSv1.0 when connecting to Google.)

We also need to remember to turn TLSv1.1 and TLSv1.2 back on when they get enabled in the next ESR...

added TorBrowserTeam201408 component::circumvention/meek owner::dcf priority::medium resolution::fixed status::closed type::defect labels

Trac:
0001-Set-security.tls.version.max-1-in-meek-http-helper.patch

0001-Set-security.tls.version.max-1-in-meek-http-helper.patch​ disables TLSv1.1 and TLSv1.2 in the helper, by setting security.tls.version.max=1.

Before the patch (i.e., the status quo with 4.0-alpha-1), the TLS fingerprint stands out from ordinary Firefox 24 in the TLS version and in an extra extension:

     SSL Record Layer: Handshake Protocol: Client Hello
         Content Type: Handshake (22)
         Version: TLS 1.0 (0x0301)
-        Length: 169
+        Length: 191
         Handshake Protocol: Client Hello
             Handshake Type: Client Hello (1)
-            Length: 165
-            Version: TLS 1.0 (0x0301)
+            Length: 187
+            Version: TLS 1.2 (0x0303)
             Random
-                gmt_unix_time: Jul 12, 2089 08:23:06.000000000 PDT
-                random_bytes: f0b149a04ac4a554c5bda57030b17342cc1c0ab59c925cc8...
+                gmt_unix_time: Nov 29, 2031 00:35:52.000000000 PST
+                random_bytes: 4856792ce5d7e72f3255fef9792ed37d14124c402ed8dfb1...
             Session ID Length: 0
             Cipher Suites Length: 70
             Cipher Suites (35 suites)
@@ -51,7 +51,7 @@
             Compression Methods Length: 1
             Compression Methods (1 method)
                 Compression Method: null (0)
-            Extensions Length: 54
+            Extensions Length: 76
             Extension: server_name
                 Type: server_name (0x0000)
                 Length: 19
@@ -86,3 +86,7 @@
             Extension: next_protocol_negotiation
                 Type: next_protocol_negotiation (0x3374)
                 Length: 0
+            Extension: signature_algorithms
+                Type: signature_algorithms (0x000d)
+                Length: 18
+                Data (18 bytes)

After the patch, we're back to differing only in the client randomness:

             Length: 165
             Version: TLS 1.0 (0x0301)
             Random
-                gmt_unix_time: Jul 12, 2089 08:23:06.000000000 PDT
-                random_bytes: f0b149a04ac4a554c5bda57030b17342cc1c0ab59c925cc8...
+                gmt_unix_time: Sep 24, 1976 08:40:40.000000000 PDT
+                random_bytes: 52240b209956653bf5fd16b29aeb040d7a81d3358f86dd19...
             Session ID Length: 0
             Cipher Suites Length: 70
             Cipher Suites (35 suites)

Trac:
Keywords: N/A deleted, TorBrowserTeam201408 added
Status: new to needs_review

I made #12873 (moved) so as not to forget to undo this change once rebased on Firefox 31.

Merged as 3c581f8426f4fd477eb26d2893dbdec42d76e23d, thanks.

Trac:
Status: needs_review to closed
Resolution: N/A to fixed

Hrmm. I think you probably want to blend in with Firefox 31 regardless. It is more common than FF24ESR. Though, I'm not sure if security.tls.version.max=3 looked like Firefox 31 if it was done from a Firefox 24 build. It may not.

Replying to mikeperry:

Hrmm. I think you probably want to blend in with Firefox 31 regardless. It is more common than FF24ESR. Though, I'm not sure if security.tls.version.max=3 looked like Firefox 31 if it was done from a Firefox 24 build. It may not.

It's a nice idea, but the ciphersuites, at least, offered by Firefox 31 are different than those offered by 24:

• https://bugzilla.mozilla.org/show_bug.cgi?id=936828
• https://bugzilla.mozilla.org/show_bug.cgi?id=946147#c1 amusing followup about broken servers ignoring the MSB of the ciphersuite ID See https://www.ssllabs.com/ssltest/viewMyClient.html for an online test. I don't think we'll win in the long run making Firefox try to look like anything other than its true version; there are too many dead-parrot issues. Firefox 24 is less common than Firefox 31, but Firefox 24 ciphersuites with TLSv1.2 is likely to be less common than both.

closed

mentioned in issue #12873 (moved)

moved to tpo/anti-censorship/pluggable-transports/meek#12766 (closed)

mentioned in issue tpo/anti-censorship/pluggable-transports/meek#12873 (closed)