tlswatch is a tool to enforce a STARTTLS policy for your mail server, combining opportunistic TLS discovery and a "sticky" trust-on-first-use policy. It currently only supports Postfix.
tlswatch will scan your email logs looking for remote mail servers that support TLS connections, and it will attempt to establish a trust path to them. If it succeeds, it will generate a new Postfix TLS policy to lock the trust path for the destination domain.
There are two available enforcement policies:
- This policy will match the exact fingerprints of the mail servers' X509 certificates. While more precise, it will generate lots of conflicts on every key rotation, so it is advised to use it only if an independent validation mechanism is available.
- This policy will match certificates against the top-level signing CA. For this to work, the remote mail server must send the full certificate chain on every TLS connection (in case the CA is not part of the set of well-known authorities distributed with the system).
It is also possible to restrict policies to a set of manually specified domains, or to provide a blacklist of domains to exclude.
Key rotation and error reconciliation
When a TLS connection error occurs, a decision must be made as to the whether change is legitimate or not. Since email to the destination domain will be queued until the condition is resolved, we'd like to attempt to validate the change automatically. A few mechanisms come to mind:
- DANE/TLSA to exploit the independent DNSSEC-based channel
- PGP web-of-trust
Manual resolution is always available as a last resort.