Distributed Computing To Stop Spam
from the we-need-something dept
The BBC has an article about a distributed computing system that works as a spam filter. It shares spam characteristics among 20 computers and looks to block out spam. Thus, if a spam message is caught on one machine – the “signature” of that message is immediately accessible to the other machines. Apparently, it works pretty well. Of course, I’m not completely sure why such a system needs to work in a distributed manner. It seems as though it could work just as easily on a centralized system.
Comments on “Distributed Computing To Stop Spam”
Reason for being distributed
I think the main reason they call it a distributed effort is because they have lots of people submitting spam. This way they get a database of lots of signatures, which makes their filtering more efficient. Sure, they could do it with just one person reporting spam, but that would result in fewer signatures and less efficient filters.
Re: Reason for being distributed
No, Ryan, I think you missed Mike’s point: centralization has its advantages. This is a classic “network” benefit application: if any one person (not just one person) labels a particular e-mail as spam, the rest of the people in the network can benefit if that tag persists (i.e., their copy of that e-mail is deleted or simply tagged and filed, any number of ways).
Mike, SpamCop essentially is a centralized clearinghouse for spam. Furthermore, they have done enough spam reporting that they now can spot a spoofed header at 50 hops, allowing them to kill spam before anyone reports it. They are also networking with latest-surviving ORBS players to kill spam from open relays. So, I’d say that subscribers reap great network benefits from the centralized service.
Not sure what distributed computing does to help this, btw, but I’ll read the article first ;).