Native TLS accepts undersized or malformed RSA certificate public keys

Summary

Reseau 1.1.0 through 1.3.1 enforce an upper bound on RSA certificate keys but no minimum modulus size and no sanity checks on the public exponent. A certificate chain containing a factorable RSA key can therefore pass Reseau's key-strength checks.

### Impact

Exploitation requires a weak RSA key to exist in a certificate chain the endpoint already trusts. An attacker who factors that key can forge certificate or TLS-handshake signatures, depending on where the weak key appears, and impersonate a peer. Exposure is primarily in private certificate authorities and legacy or misconfigured deployments; modern public certificate authorities do not normally issue sub-1024-bit RSA keys.

### Technical details

The native X.509 parser rejected RSA moduli larger than 8192 bits, but accepted arbitrarily small moduli and public exponents smaller than three or divisible by two. Those keys could then be used for certificate-chain and TLS-handshake signature verification.

This validation path was introduced with Reseau's native TLS stack in version 1.1.0. Releases 1.0.x used the previous OpenSSL TLS engine and are not affected.

### Patches

Upgrade to Reseau 1.3.2 or later. The fixed parser rejects RSA certificate keys smaller than 1024 bits and rejects public exponents that are smaller than three or even.

Applications with stronger policy requirements should independently require at least 2048-bit RSA keys; the package's patched 1024-bit floor is a compatibility baseline, not a modern key-size recommendation.

### Workarounds

Validate every RSA public key in the deployed certificate and trust chains out-of-band for an adequate modulus size and a valid odd exponent, or restrict the trust store to certificate authorities that enforce such a policy. Affected Reseau releases do not expose a configuration callback for performing this validation during a handshake. Deployments using only non-RSA certificate chains avoid this issue.

Reported to the JuliaLang security team through Anthropic's Coordinated Vulnerability Disclosure program.