Talk:Detailed project description

From The Okopipi Wiki

There is a real world problem with the heirarchy of users. If organized crime acting as a spammer decides the best way to attack this solution is to take on the people at the top, their identities would have to be very well guarded. My personal belief is that this could be ameriolated by having a trust model built upon successful categorizations. So if you've done 50+ categorizations, you're eligible to become the next tier up, and so on, like Thawte's web of trust.

The second part is that a valid PGP signature gives attackers a gold class method of attacking individuals, either through identity theft or via real world attacks. I'd really recommend just self-signed certs (SSL) or PGP without real names. The problem with using real certs is that although we're not evil, the attackers are. We cannot identify key individuals and think they will be protected by some magic method. The trust model has to be trust has to be earned, not implicit from who you are.

Random thoughts

1. It might be best to limit what a script can do to posting a specific message on the target site. This would also limit any ability to subvert the client to do other things.

2. I'd kind of like to see a two-tiered message, based on the number of participating clients in the network. At first, it would be a somewhat anonymous message, perhaps "Please stop spamming me. I'd tell you who I am, but, with the PharmaMaster/Blue Security thing, well, you understand. Please stop spamming me anyway." Who knows, maybe that would get PM in trouble. ;-(

Then, when membership passes, say, half a million, it could be more specific" "Please stop spamming xxx@yyy.zzz. Posted by Okopipi."

3. In the early stage, we might want to have an escalation method. 1st spam for a site, one optout. 2nd piece, 2 optouts. Etc. We could drop that when we reached sufficient size.

These would require a bulletproof method of tracking and passing number of active clients to all clients (or some way for distribution nodes to trigger the change).

4. It might also be helpful to send a complaint to the ISP where the email originated, complaining that such and such an email address is probably part of a botnet.

5. Is there any way we could use SpamCop to do preliminary analysis of spam for us?

6. Rather than trying to super-secure a single network, how about having multiple independant networks so that individuals can choose which set of networks to trust? In that way the project could not be destroyed by penetrating the defenses or by a single bad actor who had wormed his way into a trusted position. The project can simply publish the technology, and several operators can set it up. The reliance on keeping the core servers secret puts the entire project at too much risk of being destroyed.