Support for micro-ecc (WIP) #228

obgm · 2024-02-18T12:41:13Z

This is a proof of concept for supporting micro-ecc as a replacement for the ECC implementation in tinydtls.
micro-ecc is included as a submodule, therefore you either need to specify --recurse-submodules with git clone or do a git submodule update --init --recursive in an existing working copy.

For backwards compatibility, only curve secp256r1 is supported.

This change is based on PR #227 to avoid warnings during the build process.

Fixes issue eclipse#209. Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

The previous version ignores the option for the local port. That may be caused by issues using the same default local port for the server and client. This enables the use of an specific local port and changes the default to an ephemeral free port, similar to quite a lot of other UDP clients. The DEFAULT_PORT is therefore only used for the destination. Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

Simple client side implementation indicates support and uses the cid of the server, when negotiated by that. Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

Do not calculate the cookie using the Extensions as these are different between DTLS1.2 and DTLS1.3 https://datatracker.ietf.org/doc/html/rfc6347#section-4.2.1 When responding to a HelloVerifyRequest, the client MUST use the same parameter values (version, random, session_id, cipher_suites, compression_method) as it did in the original ClientHello. The server SHOULD use those values to generate its cookie and verify that they are correct upon cookie receipt. https://www.rfc-editor.org/rfc/rfc9147.html#section-5.3 The ClientHello up to, but not including the Extensions is the same for DTLS1.2 and DTLS1.3 Signed-off-by: Jon Shallow <supjps-libcoap@jpshallow.com>

To avoid doxygen warnings regarding outdated configuration options, this change updates the Doxyfile template to the current version 1.9.8. Change-Id: Ia60a20e4cb3da375cde48c9ae3ca234200bb8a10

The SUB_OBJS dependency rule invoked make -C on the target directory without passing important flags such as CPPFLAGS. This change removes this rule so that these objects are built with the all flags set in this Makefile. Change-Id: Ic673a881f18cbfa773860d722041a7a083f65d8e

Sets OPT_OBJS and CPPFLAGS to build uECC with curve secp256r1 Change-Id: I3c4860fbc568a492082eeb08e4ee3abd6fbc3c6e

When compiled with support for ECC, the pseudo-random number generator must be set for micro-ecc. As the function signature required by micro-ecc is different from dtls_prng(), a wrapper function is required to map the different size types. As dtls_prng's size type is larger, no other conversion is necessary. Change-Id: I2f0da18983256be3bc27f18079cf9a774049de33

As micro-ecc comes with its own unit tests, the tests for the internal ECC implementation are removed. Change-Id: I36826df4a99bd916659898587a94ea6d76af4a33

Change-Id: I85de2c3c88063dae35e81b2b3999bf7d76d04ce7

Add submodule ext/micro-ecc from https://github.com/kmackay/micro-ecc.git Change-Id: I17efa4c2fd9bd952b479bce32f0ebbb764301ca3

Change-Id: I5bd1f510c162c5ef4432f5c9cb93836312706a74

This change provides support for the curve secp256r1 from micro-ecc. Change-Id: I2d272e2ddb498016a2d6e85af7af8247010768d8

boaks · 2024-02-26T15:49:36Z

crypto.c

-  assert(sizeof(sign) >= 2 * uECC_BYTES);
+  curve_size = uECC_curve_public_key_size(uecc_curve);
+
+  assert(key_size == (unsigned int)curve_size);


That fires on my interop test. Adding

dtls_warn("%d != %d\n", (int) key_size, curve_size);

I get

Feb 26 16:46:45 WARN 32 != 64 tinydtls-server: crypto.c:638: dtls_ecdsa_verify_sig_hash2: Assertion `key_size == (unsigned int)curve_size' failed.

AFAIK, the key-size is passed in as sizeof(config->keyx.ecdsa.other_pub_x) in dtls.c, line 2589

Beside the assert, the first interop-tests with Californium are successful.
Great!

boaks · 2024-02-26T16:31:55Z

crypto.c

  uint8_t pub_key[2 * DTLS_EC_KEY_SIZE];
+  memcpy(pub_key, pub_key_x, key_size);
+  memcpy(pub_key + key_size, pub_key_y, key_size);
+  return dtls_ecdsa_verify_sig_hash2(pub_key, key_size,


I guess, the size of pub_key isn't that of the passed in key_size.

boaks · 2024-02-26T16:40:56Z

crypto.c


 #include "global.h"
 #include "dtls_debug.h"
 #include "numeric.h"
 #include "dtls.h"
 #include "crypto.h"
 #include "ccm.h"
-#include "ecc/micro-ecc/uECC.h"
+#ifdef DTLS_ECC
+#include "ext/micro-ecc/uECC.h"


"ext/micro-ecc" is part of the include path, maybe it's simpler to use "uECC.h"?

It is my preference to always include the package information path in the #include statement, to prevent name clashes (unlikely with uECC.h, but with possible with names like global.h etc.).

So, I would recommend that the include path includes ext/, and use #include <micro-ecc/uECC.h>.

The pain is, that "micro-ecc/uECC.c" uses:

#include "uECC.h" #include "uECC_vli.h"

and "micro-ecc/uECC_vli.h" uses:

#include "uECC.h" #include "types.h"

Anything that is getting built external to micro-ecc will need the equivalent of "micro-ecc/uECC.h" to get at the function definitions. For the actual micro-ecc code compilation, I assume that it is done in the micro-ecc/ directory, in which case just "uECCH.h" without a path is fine.

If you declare #include "" , the compiler first searches in current directory of source file and if not found, continues to search in any included -I directories and then the default include directories. If you declare #include <> , the compiler searches directly in any included -I directories and then the default include directories. $ cpp -v shows the include search directory order.

OK. I've checked it and at least the cmake build doesn't need to extend the include path with "ext/micro-ecc".

"ext/micro-ecc" is part of the include path,

I mixed it up, it's only part of "FILES".

boaks · 2024-02-26T17:09:19Z

See PR #229 for the fix of the "assert key_size", and support for cmake and, github workflow.

boaks · 2024-02-27T18:32:42Z

dtls.c

@@ -337,6 +349,9 @@ dtls_init(void) {
 memarray_init(&dtlscontext_storage, dtlscontext_storage_data,
              sizeof(dtls_context_t), DTLS_CONTEXT_MAX);
 #endif /* RIOT_VERSION */
+#ifdef DTLS_ECC
+ uECC_set_rng(uecc_rng_function);
+#endif /* DTLS_ECC */


Maybe that could be moved to "crypto.c". In my opinion, that would limit the scope of "ecc-related changes" to crypto.

boaks and others added 17 commits December 22, 2023 11:12

dtls.c: add check for valid handshake message type when version is 1.0.

6cf43c0

Fixes issue eclipse#209. Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

dtls_debug.h: explicitly cast macro parameter to size_t.

0f1be30

Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

dtls-server.c: check for left cli-arguments.

ff73600

Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

dtls-client.c: accept options after arguments.

44eddf4

Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

Add client dtls connection ID.

9232271

Simple client side implementation indicates support and uses the cid of the server, when negotiated by that. Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

dtls-client.c: add support_cid option.

0580b36

Signed-off-by: Achim Kraus <achim.kraus@cloudcoap.net>

[Documentation] Update Doxyfile.in to version 1.9.8

3ce86c2

To avoid doxygen warnings regarding outdated configuration options, this change updates the Doxyfile template to the current version 1.9.8. Change-Id: Ia60a20e4cb3da375cde48c9ae3ca234200bb8a10

configure.ac: Set flags for uECC if ECC support is enabled

121cad5

Sets OPT_OBJS and CPPFLAGS to build uECC with curve secp256r1 Change-Id: I3c4860fbc568a492082eeb08e4ee3abd6fbc3c6e

[Tests] Remove unit tests for internal ECC implementation

f972a67

As micro-ecc comes with its own unit tests, the tests for the internal ECC implementation are removed. Change-Id: I36826df4a99bd916659898587a94ea6d76af4a33

crypto.c: Replace ECC functions with micro-ecc

8014f05

Change-Id: I85de2c3c88063dae35e81b2b3999bf7d76d04ce7

[ECC] Add micro-ecc as submodule

f8204f2

Add submodule ext/micro-ecc from https://github.com/kmackay/micro-ecc.git Change-Id: I17efa4c2fd9bd952b479bce32f0ebbb764301ca3

Makefile.in: Add in micro-ecc dependencies

0e068ec

Change-Id: I5bd1f510c162c5ef4432f5c9cb93836312706a74

crypto.[hc]: Use micro-ecc for ECC support

3dc62b1

This change provides support for the curve secp256r1 from micro-ecc. Change-Id: I2d272e2ddb498016a2d6e85af7af8247010768d8

obgm mentioned this pull request Feb 18, 2024

Experiencing an extended handshake duration on a SAMR21-xpro device. #224

Open

boaks reviewed Feb 26, 2024

View reviewed changes

boaks reviewed Feb 27, 2024

View reviewed changes

boaks force-pushed the develop branch 2 times, most recently from 78f3535 to 0d3aa09 Compare May 13, 2024 09:48

boaks force-pushed the develop branch from 0d3aa09 to 039e6e2 Compare June 2, 2024 15:22

boaks force-pushed the develop branch 2 times, most recently from c6dd8d8 to db5b584 Compare August 28, 2024 14:30

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support for micro-ecc (WIP) #228

Support for micro-ecc (WIP) #228

obgm commented Feb 18, 2024

boaks Feb 26, 2024

boaks Feb 26, 2024

boaks Feb 26, 2024

boaks Feb 26, 2024

mrdeep1 Feb 26, 2024 •

edited

Loading

boaks Feb 26, 2024

mrdeep1 Feb 26, 2024

boaks Feb 27, 2024

boaks Feb 27, 2024

boaks commented Feb 26, 2024

boaks Feb 27, 2024

Support for micro-ecc (WIP) #228

Are you sure you want to change the base?

Support for micro-ecc (WIP) #228

Conversation

obgm commented Feb 18, 2024

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

mrdeep1 Feb 26, 2024 • 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

boaks commented Feb 26, 2024

Choose a reason for hiding this comment

mrdeep1 Feb 26, 2024 •

edited

Loading