CVE-2026-41509 is a critical buffer overflow (CVSS 9.8) in the CROSS post-quantum signature implementation. In crypto_sign_open(), an integer underflow of the message length leads to the overflow. The fix is commit fc6b7e7. Update any build that includes this code.
What's affected
Per the record, builds of the CROSS implementation before commit fc6b7e7 are affected. The GitHub advisory lists no patched release, only that commit. It also says the report was made to liboqs on March 7, 2026, and calls the bug an implementation flaw that is not related to the theoretical security of the signature scheme.
Check whether your software vendors the CROSS code directly or through a library such as liboqs, since a copy of the source may sit inside a larger project.
How to check if you're exposed
Search your source trees and submodules for CROSS signature code:
grep -R "crypto_sign_open" --include=*.c --include=*.h . | head
Check the git history of any vendored copy for commit fc6b7e7. If it is missing, you have the old code. For a library that bundles it, read that project's release notes for the CROSS fix.
Also work out whether the verification function handles input you do not control, such as signed messages from other parties.
What to do now
- Update the CROSS code to include commit fc6b7e7, or move to a release that does.
- If you consume it through another library, look for that library's patched release and upgrade.
- Rebuild and redeploy everything that links the code.
- Until then, avoid verifying signatures on messages from untrusted sources with the vulnerable function.
Why it matters
The record lists CWE-121 and CWE-122, stack based and heap based buffer overflow. The advisory ties the flaw to an underflow of the integer mlen. When a length that should be positive wraps to a huge value, later copies can write far past a buffer.
Post-quantum cryptography is new code written in C, and implementation bugs are as likely there as anywhere. The math can be sound while the code that runs it is not. The record does not flag this CVE as known exploited.
Quick check
Does this mean the CROSS scheme is broken?
No. The advisory says it is an implementation vulnerability, not related to the theoretical security of the scheme.
Sources
Last reviewed: 19 October 2026.
