s2n-tls has a potentially observable differences in RSA premaster secret handling

When receiving a message from a client that sent an invalid RSA premaster secret, an issue in s2n-tls results in the server performing additional processing when the premaster secret contains an incorrect client hello version. While no practical attack on s2n-tls has been demonstrated, this causes a small timing difference which could theoretically be used as described in the Marvin Attack [1].

We would like to thank Hubert Kario [2] for reporting this issue.

Impact

The extent of this issue is a timing difference. No practical attack on s2n-tls has been demonstrated.

This issue affects server applications that permit RSA key exchange. Applications that use the default, built-in blinding feature or properly implement self-service blinding are not affected.

Impacted versions: <= v1.4.15.

Patches

The patch is included in v1.4.16 [3].

Workarounds

Applications can work around this issue by using an s2n-tls security policy that disallows RSA key exchange.

If you have any questions or comments about this advisory we ask that you contact AWS/Amazon Security via our vulnerability reporting page [4] or directly via email to [email protected]. Please do not create a public GitHub issue.

[1] https://people.redhat.com/~hkario/marvin/
[2] https://github.com/tomato42
[3] https://github.com/aws/s2n-tls/releases/tag/v1.4.16
[4] https://aws.amazon.com/security/vulnerability-reporting