[GHSA-wj64-gh9j-xm82] Issue summary: An OpenSSL TLS 1.3 server may fail to...

advisories/unreviewed/2026/03/GHSA-wj64-gh9j-xm82/GHSA-wj64-gh9j-xm82.json

@@ -1,19 +1,37 @@
 {
   "schema_version": "1.4.0",
   "id": "GHSA-wj64-gh9j-xm82",
-  "modified": "2026-03-17T18:30:32Z",
+  "modified": "2026-03-17T18:31:38Z",
   "published": "2026-03-13T21:31:46Z",
   "aliases": [
     "CVE-2026-2673"
   ],
-  "details": "Issue summary: An OpenSSL TLS 1.3 server may fail to negotiate the expected\npreferred key exchange group when its key exchange group configuration includes\nthe default by using the 'DEFAULT' keyword.\n\nImpact summary: A less preferred key exchange may be used even when a more\npreferred group is supported by both client and server, if the group\nwas not included among the client's initial predicated keyshares.\nThis will sometimes be the case with the new hybrid post-quantum groups,\nif the client chooses to defer their use until specifically requested by\nthe server.\n\nIf an OpenSSL TLS 1.3 server's configuration uses the 'DEFAULT' keyword to\ninterpolate the built-in default group list into its own configuration, perhaps\nadding or removing specific elements, then an implementation defect causes the\n'DEFAULT' list to lose its 'tuple' structure, and all server-supported groups\nwere treated as a single sufficiently secure 'tuple', with the server not\nsending a Hello Retry Request (HRR) even when a group in a more preferred tuple\nwas mutually supported.\n\nAs a result, the client and server might fail to negotiate a mutually supported\npost-quantum key agreement group, such as 'X25519MLKEM768', if the client's\nconfiguration results in only 'classical' groups (such as 'X25519' being the\nonly ones in the client's initial keyshare prediction).\n\nOpenSSL 3.5 and later support a new syntax for selecting the most preferred TLS\n1.3 key agreement group on TLS servers.  The old syntax had a single 'flat'\nlist of groups, and treated all the supported groups as sufficiently secure.\nIf any of the keyshares predicted by the client were supported by the server\nthe most preferred among these was selected, even if other groups supported by\nthe client, but not included in the list of predicted keyshares would have been\nmore preferred, if included.\n\nThe new syntax partitions the groups into distinct 'tuples' of roughly\nequivalent security.  Within each tuple the most preferred group included among\nthe client's predicted keyshares is chosen, but if the client supports a group\nfrom a more preferred tuple, but did not predict any corresponding keyshares,\nthe server will ask the client to retry the ClientHello (by issuing a Hello\nRetry Request or HRR) with the most preferred mutually supported group.\n\nThe above works as expected when the server's configuration uses the built-in\ndefault group list, or explicitly defines its own list by directly defining the\nvarious desired groups and group 'tuples'.\n\nNo OpenSSL FIPS modules are affected by this issue, the code in question lies\noutside the FIPS boundary.\n\nOpenSSL 3.6 and 3.5 are vulnerable to this issue.\n\nOpenSSL 3.6 users should upgrade to OpenSSL 3.6.2 once it is released.\nOpenSSL 3.5 users should upgrade to OpenSSL 3.5.6 once it is released.\n\nOpenSSL 3.4, 3.3, 3.0, 1.0.2 and 1.1.1 are not affected by this issue.",
+  "summary": "In rare configurations an OpenSSL 3.5/3.6 server may select a non-PQC client keyshare instead of HRR",
+  "details": "Issue summary: An OpenSSL TLS 1.3 server may fail to negotiate the expected\npreferred key exchange group when its key exchange group configuration includes\nthe default by using the 'DEFAULT' keyword.\n\nImpact summary: A less preferred key exchange may be used even when a more\npreferred group is supported by both client and server, if the group\nwas not included among the client's initial predicated keyshares.\nThis will sometimes be the case with the new hybrid post-quantum groups,\nif the client chooses to defer their use until specifically requested by\nthe server.\n\nIf an OpenSSL TLS 1.3 server's configuration uses the 'DEFAULT' keyword to\ninterpolate the built-in default group list into its own configuration, perhaps\nadding or removing specific elements, then an implementation defect causes the\n'DEFAULT' list to lose its 'tuple' structure, and all server-supported groups\nwere treated as a single sufficiently secure 'tuple', with the server not\nsending a Hello Retry Request (HRR) even when a group in a more preferred tuple\nwas mutually supported.\n\nAs a result, the client and server might fail to negotiate a mutually supported\npost-quantum key agreement group, such as 'X25519MLKEM768', if the client's\nconfiguration results in only 'classical' groups (such as 'X25519' being the\nonly ones in the client's initial keyshare prediction).\n\nOpenSSL 3.5 and later support a new syntax for selecting the most preferred TLS\n1.3 key agreement group on TLS servers.  The old syntax had a single 'flat'\nlist of groups, and treated all the supported groups as sufficiently secure.\nIf any of the keyshares predicted by the client were supported by the server\nthe most preferred among these was selected, even if other groups supported by\nthe client, but not included in the list of predicted keyshares would have been\nmore preferred, if included.\n\nThe new syntax partitions the groups into distinct 'tuples' of roughly\nequivalent security.  Within each tuple the most preferred group included among\nthe client's predicted keyshares is chosen, but if the client supports a group\nfrom a more preferred tuple, but did not predict any corresponding keyshares,\nthe server will ask the client to retry the ClientHello (by issuing a Hello\nRetry Request or HRR) with the most preferred mutually supported group.\n\nThe above works as expected when the server's configuration uses the built-in\ndefault group list, or explicitly defines its own list by directly defining the\nvarious desired groups and group 'tuples'.\n\nThis is not a common configuration, the `DEFAULT` keyword is new in this context,\nand not many servers are using it. Both the client and the server have to support\none of the new PQC key exchange algorithms, and the client (also non-default\nbehaviour) has to choose to defer its PQC keyshare until a server requests it\nthrough HRR.  The impact is low, previous OpenSSL versions always accept\na mutually supported keyshare without HRR even if the client could have sent\na more preferred variant (all key exchange algorithms were considered\n*sufficiently* secure). The client is clearly willing to use a non-PQC keyshare,\nwhich is what it sends, leaving it to the server to optionally suggest something\nstronger.  The OpenSSL team considers this a **LOW** severity issue.\n\nThis is a configuration issue, there is no attack vector, the server does not respond\nto the client the way that the server's administrator may have intended.\n\nNo OpenSSL FIPS modules are affected by this issue, the code in question lies\noutside the FIPS boundary.\n\nOpenSSL 3.6 and 3.5 are vulnerable to this issue.\n\nOpenSSL 3.6 users should upgrade to OpenSSL 3.6.2 once it is released.\nOpenSSL 3.5 users should upgrade to OpenSSL 3.5.6 once it is released.\n\nOpenSSL 3.4, 3.3, 3.0, 1.0.2 and 1.1.1 are not affected by this issue.",
   "severity": [
     {
       "type": "CVSS_V3",
-      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N"
+      "score": "CVSS:3.1/AV:L/AC:H/PR:H/UI:R/S:U/C:L/I:N/A:N"
+    }
+  ],
+  "affected": [
+    {
+      "package": {
+        "ecosystem": "Pub",
+        "name": ""
+      },
+      "ranges": [
+        {
+          "type": "ECOSYSTEM",
+          "events": [
+            {
+              "introduced": "0"
+            }
+          ]
+        }
+      ]
     }
   ],
-  "affected": [],
   "references": [
     {
       "type": "ADVISORY",
@@ -40,7 +58,7 @@
     "cwe_ids": [
       "CWE-757"
     ],
-    "severity": "HIGH",
+    "severity": "LOW",
     "github_reviewed": false,
     "github_reviewed_at": null,
     "nvd_published_at": "2026-03-13T19:54:34Z"