4 ms·
<meta property="og:description" content="Search the web privately ..."> Would be nice if Brave did not require SNI since this is considered a privacy concern
by 1vuio0pswjnm7 4y ago
<meta property="og:description" content="Search the web privately ...">
Would be nice if Brave did not require SNI since this is considered a privacy concern by some folks.^1 Anyone sniffing the wire can see all the domain names to which the SNI user is connecting.^3 The other search engines do not require SNI, e.g., Google, Bing, DuckDuckGo, GigaBlast, Qwant, etc.
1. One example would be Cloudflare. Because some folks see SNI as a privacy concern, Cloudflare used to offer ESNI which was a way to encrypt SNI. It has since been discontinued while we wait for ECH. Some HN commenters will often try to argue that SNI is irrelevant to users without offering any evidence to support. Watch for it. For example, China found SNI was relevant enough to block ESNI. Apparently, China found it preferable to use SNI than to use only IP addresses, which of course are easy for websites to change. Go figure.
https://mailarchive.ietf.org/arch/msg/tls/Dae-cukKMqfzmTT4Ksh1Bzlx7ws/ https://mailarchive.ietf.org/arch/msg/tls/Dae-cukKMqfzmTT4Ks...
SNI can be used for censorship purposes, among other things. Many search engines work without SNI. But not Brave.
NB. As I understand it, these browsers do not allow the user to enable/disable SNI on a per site basis; in some of them it is not even possible to disable SNI at all.^2 TLS might enable the user to hide web pages from the proverbial "MITM", but with SNI enabled it will not allow them to hide web sites.
2. Thus, even when Google, Bing, DuckDuckGo, GigaBlast, Qwant, Mojeek, etc., and millions of other websites do not require the user to send SNI in order to return SERPs or other pages, these browsers send it anyway. Brilliant.
3. SNI is different than DNS. DNS lookups can be done at a different time from when a user connects, if the user ever does connect. (Popular browsers are not good for this, of course.) Unlike DNS, SNI proves the user actually connected. Strangely, much effort has gone into encrypting DNS, while SNI, and to some extent TLS prior to version 1.3, leaks these same domain names on the wire, unencrypted.
- 1vuio0pswjnm7 4y agoTruly, use of AWS is the culprit here since it requires SNI. Perhaps Brave will one day graduate from using a large third party CDN.
- josephcsible 4y agoIf you can choose not to require SNI, doesn't that mean that your domain is the only one being served by the IP it's on, and that the IP address (also visible to anyone sniffing the wire) then reveals the exact same information as SNI would have?
- 1vuio0pswjnm7 4y agoNot necessarily. For example, if I access 142.250.187.228 what domain name was in the Host header. Hint: It's not www.google.com. If I visit 65.9.141.10, then what domain was in the Host header. Hint: It's not search.brave.com. But do not take a random HN commenter's word for it. Try this yourself. Try to map IP addresses to domain names on a very large sample. I have tested this on hundreds of thousands of domains. It will not work reliably. Not even close. Compared to SNI, which is both simple and reliable, it is a total PITA. If this 1-to-1 IP to domain mapping and 100% reliable reverse DNS idea made any sense, then why claim encrypted DNS offers "DNS privacy". Anyone can see IP addresses on the wire. If all IP addresses can easily be mapped to domain names, then how would this provide any privacy. Mapping IP addresses to domain names has to account for the fact that IP addresses can change. To make sure the mapping is correct one has to constantly keep doing lookups. SNI stays the same even if the IP changes. It requires very little effort; it makes the process of determining the domain name trivial. The goal is not total privacy. (Impossible when using the internet anyway.) The goal is to not to make surveillance easy. That's what SNI does. It makes surveillance (and censorship) super easy. If I am wrong about this, and SNI is AOK, then why was ESNi developed and why is ECH being developed. It is not an easy problem to solve. Why bother. SNI was a feature to benefit websites, not website users. It has some very undesirable properties for users. Even with CDNs that need to use SNI, it is still possible to avoid sending the domain name in plaintext on the wire. One can use a dummy name, i.e., not the name in the Host header, as the SNI and the CDN can still return the correct page because it uses only the Host header. This was nicknamed "domain fronting". I was using this technique before anyone started calling it that. Using a dummy name can reduce the surveillance value of SNI, sometimes to zero. After some "security researchers" started bragging about their discovery of "domain fronting", AWS and other CDNs started checking SNI against Host header to make sure they are the same. Now "domain fronting" no longer works. Countries that censor the internet, who are valued AWS customers, rely on accurate SNI in order to carry out censorship. Go figure.