> It has the side effect of sending every domain you visit to the certificate log server.
No? It means that if a certificate is to be trusted, it needs to enter a public append only log server.
It does not "send every domain you visit to CT." The certificate simply has to be in CT before chrome trusts it.
CT logs are public, you can (and many do) make archives of it.
The idea is simple, if a website is to be _publicly trusted_, it needs to also be publicly logged.
You have heavily misunderstood what CT does, and the paradigm of control around it.
You're not thinking about it hard enough. Your CT's would effectively have to be implemented in a way where everyone had a locally cached copy of the log to check against. That at least shuts down an browsing activity leak vector.
If you centralize it, the abuse is a matter of when, not if.
Again, misunderstanding of how it works.
Logs are checked using the SCT in the certificate. It's a signed "promise" from the CT that the pre-certificate will eventually be appended to the logs in a few hours since submission.
The various root programs monitor to ensure that the promise does in fact happen.
Your browser process does *not* reach out to the CT log on its own.
I've worked at both Let's Encrypt and Google Trust Services. CT is something I'm intimately familiar with.
> it has to check the cert log.
No, it does not.
> Yes but how does your browser trust it?
Your browser ships with a log list: https://source.chromium.org/chromium/chromium/src/+/main:com...
The public key of the log is included there. It's checked to see if the SCT is signed with the correct public key.
There are three possible states:
No Safe Browsing protections -> no SCT auditing
Default Safe Browsing protections -> SCT auditing logic selects a small proportion of TLS connections and performs a k-anonymous lookup on an SCT. If that privacy-preserving SCT lookup reveals that the SCT is not known to Google but should be, the client uploads the certificate, SCTs, and hostname to Google (but no other information).
Enhanced Safe Browsing protections -> SCT auditing logic selects a small proportion of TLS connections and uploads the certificate, SCTs, and hostname to Google (but no other information).
https://groups.google.com/a/chromium.org/g/ct-policy/c/Fddjj...I was talking about the validation side. SCT Auditing basically helps ensure that the logs are keeping their promise by letting the root program (Chrome, and Apple) see what certificates are flowing on the internet and if the merkle tree structure is valid.
CT does not require that, that was part of the explicit design goal of CT. To be trusted by a browser a CA has to publish every certificate they issue to multiple CT logs. When the certificate is published to a CT log, the log provides a signed confirmation, and that is included in the final certificate.
Every browser has a set of CT logs that they trust. When they perform the trust evaluation on a certificate they verify the certificate is correctly signed by the CA, and then they go through the included CT annotations and find the annotations from logs that the browser trusts and verify that those annotations are signed with the respective log's public key.
At no point does the browser perform any other network requests (again, a problem with OCSP is that the additional network access during TLS handshakes was incredibly slow and unreliable - if CT required this it would require _more_ fallible network loads). That makes CT faster, and resolves the privacy hole OCSP forced.
https://en.wikipedia.org/wiki/Certificate_Transparency