![](/static/253f0d9b/assets/icons/icon-96x96.png)
![](https://fry.gs/pictrs/image/c6832070-8625-4688-b9e5-5d519541e092.png)
I’m not sure if it’s part of a TLS standard yet but I was talking about encrypted SNI (ECH, formerly called ESNI).
Today, early on in a TLS connection, the client actually tells the server, in plain text, the domain name it’s intending to communicate with. The server then presents a response that only the owner of that domain can produce, then keys are exchanged and the connection progresses, encrypted. This was required to allow a single server to serve traffic on multiple domains. Before this, a server on an IP:Port combo could only serve traffic on a single domain.
But because of this, a man in the middle can just read the ClientHello and learn the domain you’re intending to connect to. They can’t intercept any encapsulated data (e.g. at the HTTP level, in the case of web traffic) but they can learn the domains you’re accessing.
ECH promises to make the real ClientHello encrypted by proceeding it with a fake ClientHello. The response will contain enough information to fetch a key that can be used to encrypt the real ClientHello. Only the server will be able to decrypt this.
Whoever made this has never used Google Cloud Platform.