Hello,

CERT/CC published three vulnerability notes on 13 and 14 October 2026 about how open-source TLS/X.509 libraries enforce Name Constraints (RFC 5280 section 4.2.1.10). I reported the findings. This is a summary of public fixes and statuses; it does not announce new CVEs.

https://kb.cert.org/vuls/id/573195 https://kb.cert.org/vuls/id/948176 https://kb.cert.org/vuls/id/410595

The differences fall into four areas: DNS-name comparison (RFC 5280 requires case-insensitive matching), IP addresses (the RFC does not say whether IPv4-mapped IPv6 should be treated as IPv4), URI-authority parsing, and wildcards (the RFC does not define wildcard SAN semantics).

Fixed or merged:

pyca/cryptography CVE-2026-34073, fixed in 46.0.6. DNS constraints were checked against the certificate's SANs but not against the requested peer name. GHSA-m959-cc7f-wv43 rustls/webpki wildcards under permitted subtrees: CVE-2026-93601, fixed in 0.103.12 / 0.104.0-alpha.6 (GHSA-xgp8-3hg3-c2mh). Unsupported URI constraints were accepted: CVE-2026-93600, first fix in 0.103.12 / 0.104.0-alpha.6, excludedSubtrees case completed in 0.103.13 / 0.104.0-alpha.7 (GHSA-965h-392x-2mh5). GnuTLS CVE-2026-3833, fixed in 3.8.13. DNS names and the domain part of rfc822Name constraints compared case-sensitively. Independently co-reported. wolfSSL CVE-2026-5263, fixed in 5.9.1: URI constraints parsed but not enforced. Fragment wildcards (e.g. foo.com): PR 9991, in 5.9.1. Note: VU#948176 cites CVE-2026-3547 for the URI report; that CVE record describes an ALPN parsing issue. AWS-LC wildcard Common Name fallback fixed in 1.71.0 / aws-lc-sys 0.39.0 (GHSA-3jrg-j22w-mpmc, GHSA-394x-vwmw-crm3). The same advisory covers Unicode CN; VU#948176 lists that report as still under coordination. Go crypto/x509 CVE-2026-33810, fixed in Go 1.26.2 (GO-2026-4866, co-credited): case handling in wildcard parent-constraint lookup. OpenSSL URI authority: the userinfo scan for @ ran past /, ? or #. PR #30319; OpenSSL confirmed on 2026-09-10 that it is in all then-supported versions. BoringSSL VU#410595.8 (wildcard vs excluded host) and .12 (trailing-dot DNS names) fixed upstream in commit 5774eca (2026-03-04), legacy crypto/x509 path.

Open, disputed or declined (per the notes):

Go two reports in triage; IPv4-mapped IPv6 report not accepted as meeting the security bar OpenSSL userinfo-colon report in triage BoringSSL .9 accepted (upstream change b84dc60, 2026-05-05), .10 disputed, .11 declined Mozilla NSS IP-address report under coordination LibreSSL wildcard report under coordination s2n-tls Common Name fallback report in triage Mbed TLS non-critical Name Constraints not enforced in the reported configuration; vendor treats support as a feature request, reporter's concern disputed

Conditions, according to CERT/CC: an attacker needs a trusted signature over chosen certificate fields, an affected validation path must be reachable, and the application must use the accepted name for authorization. The notes do not report exploitation. Using one of these libraries does not by itself mean a product reaches an affected path.

No test vectors are published for anything not fixed.

Per-finding status with dated sources: https://paperlock.org/findings/

Regards, Oleh Konko 1seal - https://1seal.org/research/

In Mbed TLS 3.2.0 though 3.6.6 and 4.0.0 through 4.1.0, an attacker who can cause an entropy source to fail can remove or inject bytes into the start of the TLS stream. This only affects TLS 1.3 servers.

In Mbed TLS 3.6.x before 3.6.7 and 4.1.x before 4.1.2, the TLS 1.3 client accepts HelloRetryRequest selecting an unadvertised group.

Mbed TLS through 3.6.4 has an Observable Timing Discrepancy.

Mbed TLS before 3.6.4 has a NULL pointer dereference because mbedtlsasn1storenameddata can trigger conflicting data with val.p of NULL but val.len greater than zero.

In Mbed TLS 3.6.1 through 3.6.3 before 3.6.4, a timing discrepancy in block cipher padding removal allows an attacker to recover the plaintext when PKCS#7 padding mode is used.

Mbed TLS before 2.28.10 and 3.x before 3.6.3, in some cases of failed memory allocation or hardware errors, uses uninitialized stack memory to compose the TLS Finished message, potentially leading to authentication bypasses such as replays.

An issue was discovered in Mbed TLS 3.6 before 3.6.1. A stack buffer overflow in mbedtlsecdsadertoraw() and mbedtlsecdsarawtoder() can occur when the bits parameter is larger than the largest supported curve. In some configurations with PSA disabled, all values of bits are affected. (This never happens in internal library calls, but can affect applications that call these functions directly.)

An issue was discovered in Mbed TLS 3.5.x before 3.6.0. When an SSL context was reset with the mbedtlssslsessionreset() API, the maximum TLS version to be negotiated was not restored to the configured one. An attacker was able to prevent an Mbed TLS server from establishing any TLS 1.3 connection, potentially resulting in a Denial of Service or forced version downgrade from TLS 1.3 to TLS 1.2.

In Mbed TLS 3.3.0 through 3.5.2 before 3.6.0, a malicious client can cause information disclosure or a denial of service because of a stack buffer over-read (of less than 256 bytes) in a TLS 1.3 server via a TLS 3.1 ClientHello.