Runtime check for EC keys in Secure Session #693

iamnotacake · 2020-08-11T11:34:48Z

Deny non-EC keys when creating/negotiating Secure Session.
It was possible to create Secure Session by passing private RSA key to
secure_session_create(). But then it would fail while negotiating
connection with another peer since Secure Session doesn't actually
support RSA keys. This PR make it fail a little bit earlier.
Move test keys to different file
Add tests to make sure secure_session_create fails with EC pub key, RSA priv/pub keys.
Add test to make sure secure_session_create fail with empty peer ID (id == NULL || id_len == 0).
Mention the key limitation in changelog.

Checklist

Change is covered by automated tests
The coding guidelines are followed
Changelog is updated (in case of notable or breaking changes)

* Deny non-EC keys when creating/negotiating Secure Session. It was possible to create Secure Session by passing private RSA key to `secure_session_create`. But then it would fail while negotiating connection with another peer since Secure Session doesn't actually support RSA keys. This commit make it fail a little bit earlier. * Add test to make sure `secure_session_create` fail with RSA key. * Add test to make sure `secure_session_create` fail with empty peer ID (peer_id_len == 0).

tests/themis/themis_secure_session.c

* Move test keys to different file, make `clang-tidy` satisfied * Rewrite messages in test functions

CHANGELOG.md

src/themis/secure_session_peer.c

vixentael · 2020-08-11T16:34:53Z

tests/themis/themis_secure_session.c

-    0x1f, 0xe9, 0xea, 0x48, 0x11, 0xe1, 0xf9, 0x71, 0x8e, 0x24, 0x11, 0xcb, 0xfd, 0xc0, 0xa3,
-    0x6e, 0xd6, 0xac, 0x88, 0xb6, 0x44, 0xc2, 0x9a, 0x24, 0x84, 0xee, 0x50, 0x4c, 0x3e, 0xa0,
-};
+/* Test keys can be found in themis_secure_session_test_keys.h */


vixentael · 2020-08-11T16:35:41Z

tests/themis/themis_secure_session.c

@@ -668,11 +666,40 @@ static void test_basic_flow_no_transport(void)
    }
 }

+static void test_empty_peer_id(void)
+{
+    memcpy(&(client.transport), &transport, sizeof(secure_session_user_callbacks_t));


remember: using memcpy is ok only for tests / non-sensitive data

vixentael · 2020-08-11T16:36:10Z

tests/themis/themis_secure_session.c

+    testsuite_fail_if(client.session != NULL, "secure_session_create must fail with empty Peer ID");
+}
+
+static void test_invalid_private_key_type(void)


Let's test public key and make sure that tests are passing and write comment why is that.

IMO, it's would be better to split this into separate tests: RSA private keys, RSA public keys, EC public keys.

* Mention GitHub issue # in changelog. * Add one more check to `secure_session.c` to only allow EC private keys in `secure_session_create()`. * Add more tests to make sure the session cannot be created with public EC and public RSA keys.

CHANGELOG.md

src/themis/secure_session.c

src/themis/secure_session_peer.c

ilammy · 2020-08-12T07:51:07Z

tests/themis/themis_secure_session.c

+    memcpy(&(client_rsa.transport), &transport, sizeof(secure_session_user_callbacks_t));
+    client_rsa.transport.user_data = &client_rsa;


Well, this a weird way to setup a Secure Session – with global variables and memcpy() – but I'll let it pass given that all tests here are like that. It's not nice coding practice because it makes impossible to run the tests in parallel. However, refactoring this is a topic for another day.

I guess the easiest way to refactor this is to mark those glogal vars as thread local and then create as many threads as we like, so they will run in parallel. Another question is whether the test framework is ready for such changes. And these tests are really fast, like almost instant, and the only computation-heavy check is soter rand: NIST STS, which, btw, only uses one CPU core.

You can also just create a new client context and callbacks for each test :) Which is arguably more readable and obvious than thread-local variables. It's more verbose, yeah, but dumb test code is better than smart one, IMO.

ilammy · 2020-08-12T07:52:32Z

tests/themis/themis_secure_session.c

+    testsuite_fail_if(client.session != NULL, "secure_session_create must fail with empty Peer ID");
+}
+
+static void test_invalid_private_key_type(void)


IMO, it's would be better to split this into separate tests: RSA private keys, RSA public keys, EC public keys.

tests/themis/themis_secure_session.c

* Changed text in changelog * Removed unneeded checks: - `themis_get_asym_key_kind()` already checks pointer and length - non-EC keys won't reach `secure_session_peer_init()` so we actually don't need that check * Split tests into more funcs

ilammy

Nice work 👍

ilammy · 2020-08-12T10:06:16Z

src/themis/secure_session_peer.c

 #include "soter/soter_wipe.h"
+#include "themis/secure_keygen.h"


It guess it's not necessary to include this any more.

Forgot to remove in prev commit

iamnotacake added 2 commits August 11, 2020 12:56

Update changelog

ffd3ae5

iamnotacake added core Themis Core written in C, its packages tests Themis test suite labels Aug 11, 2020

iamnotacake changed the title ~~Anatolii t1739 runtime check ec keys in secure session~~ Runtime check for EC keys in Secure Session Aug 11, 2020

Fix test compilation issues

b8cf3af

ilammy reviewed Aug 11, 2020

View reviewed changes

tests/themis/themis_secure_session.c Outdated Show resolved Hide resolved

Fix code formatting

668db5e

* Move test keys to different file, make `clang-tidy` satisfied * Rewrite messages in test functions

vixentael reviewed Aug 11, 2020

View reviewed changes

Updates after review

2222185

* Mention GitHub issue # in changelog. * Add one more check to `secure_session.c` to only allow EC private keys in `secure_session_create()`. * Add more tests to make sure the session cannot be created with public EC and public RSA keys.

iamnotacake marked this pull request as ready for review August 12, 2020 06:50

iamnotacake requested review from ignatk, Lagovas and storojs72 as code owners August 12, 2020 06:50

ilammy reviewed Aug 12, 2020

View reviewed changes

Updates after review

b39cdfd

* Changed text in changelog * Removed unneeded checks: - `themis_get_asym_key_kind()` already checks pointer and length - non-EC keys won't reach `secure_session_peer_init()` so we actually don't need that check * Split tests into more funcs

ilammy approved these changes Aug 12, 2020

View reviewed changes

Remove unused include

4e2d4e5

Forgot to remove in prev commit

iamnotacake merged commit bef80bc into cossacklabs:master Aug 12, 2020

iamnotacake deleted the anatolii-T1739-runtime-check-ec-keys-in-secure-session branch August 12, 2020 13:39

ilammy mentioned this pull request Aug 17, 2020

Start phasing out CircleCI #709

Merged

2 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Runtime check for EC keys in Secure Session #693

Runtime check for EC keys in Secure Session #693

iamnotacake commented Aug 11, 2020 •

edited

Loading

vixentael Aug 11, 2020

vixentael Aug 11, 2020

vixentael Aug 11, 2020 •

edited

Loading

ilammy Aug 12, 2020

vixentael Aug 12, 2020

ilammy Aug 12, 2020

iamnotacake Aug 12, 2020

ilammy Aug 12, 2020

ilammy Aug 12, 2020

ilammy left a comment

ilammy Aug 12, 2020

		memcpy(&(client_rsa.transport), &transport, sizeof(secure_session_user_callbacks_t));
		client_rsa.transport.user_data = &client_rsa;

		#include "soter/soter_wipe.h"
		#include "themis/secure_keygen.h"

Runtime check for EC keys in Secure Session #693

Runtime check for EC keys in Secure Session #693

Conversation

iamnotacake commented Aug 11, 2020 • edited Loading

Checklist

Choose a reason for hiding this comment

Choose a reason for hiding this comment

vixentael Aug 11, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ilammy left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

iamnotacake commented Aug 11, 2020 •

edited

Loading

vixentael Aug 11, 2020 •

edited

Loading