I hesitate to mention it, because I had a bad experience with them and it involves "crypto," which is always a hot-button, but https://presearch.com is also playing in the distributed search engine space, and their approach actually seems reasonably well thought out <https://www.presearch.io/vision.pdf>. Unfortunately that paper does not seem to address your rogue nodes observation, which makes me doubt their success beyond their availability issues when I tried it. All of that is above and beyond the "download this closed source binary, trust us!" which noped me right out from considering running a node