FOAF + OpenID is a semantic web attempt to solve the problem of blog spam. The idea is that those people who have FOAF files and OpenID identities can identify each others networks of friends, colleagues and acquaintances using their FOAF files and authenticate the individuals using OpenID.
I’ve found a flaw in this: I could have (and probably should have) a link in my FOAF file to a semantic wiki representation of myself, which is (in the way of wikis) world writeable. Spammers could easily edit the wiki to insert a link from myself to them which would let them become part of the group and spam us.
There are a number of fixes for this:
- Check the metadata in each FOAF file to ensure that it claims to be written by the subject of the file (which wouldn’t be the case for the wiki). This would require many FOAF/RDF generation tools to be updated.
- Add trust attributes to external links in FOAF files. This would also require many FOAF/RDF generation tools to be updated.
- Compile a list of known world-writable RDF sources and use it to black-list them. This would always be playing a game of catch-up and there some sites might slip through.
- Require trusted users not to link to world-writeable RDF sources (or sources of RDF that harvest from the wider web). This requires that the semantic web workers work in a walled garden and not link outside it into the wider web.
None of these are easy.
Somehow this whole thing reminds me of the OpenPGP web-of-trust, without the cryptographic underpinnings.