Using SearX hosted by someone else is marginally better, but now you have to trust the owner of the server, which is probably not what you want for privacy-centered search engine.
Using SearX hosted by someone else is marginally better, but now you have to trust the owner of the server, which is probably not what you want for privacy-centered search engine.
Is the concern that the latter's IP isn't behind a NAT, and therefore is more unique? If so, I think that's the least concerning of the identifying datapoints that a search engine has access to -- my browser metadata is far more identifying. With SearX, that information doesn't get forwarded (IIUC).
* Share your instance with some friends (though of course it shifts the trust one level down)
* Route outgoing requests through tor / some VPN. In a hosted environment that you configure once for "everywhere", it's more feasible to do fun things like "Google searches over OperaVPN, bing searches through Mullvad, everything else through Tor". You could even change proxies.
Especially with the latter you can kind of "eat the cake and have it too", with some added latency of course.
As a bonus, searx can also be configured to rewrite links (yt->piped, twitter->nitter, imgur->rimgu etc) and remove tracking query parameters.
https://www.vultr.com/docs/install-searx-with-nginx-on-ubunt...
https://www.linode.com/content/maximize-your-privacy-with-se...
Either way, what we're talking about here is more akin to an http proxy than scanning/crawling - since every request is explicitly triggered by manual user action as opposed to automated so it shouldn't be any issue.
If you expose and advertise a public unauthenticated frontend and end up with 10ks of users maaaaybe some providers will start talking to you about it but otherwise I wouldn't have any concerns. And if you get to that scale, you may already want to look at proxying through providers like Luminati anyway.
You just reveal your search queries to the hosting provider if he maliciously intercepts them.