Peer to Peer sharing of bogofilter database information
Fast Bayesian spam filter along lines suggested by Paul Graham
Brought to you by:
m-a
It would be nice to have the capability to share
bogofilter information among a group of trusted peers.
Naturally, this capability would have to be limited
for it to be effective. (i.e., it couldn't be made a
public internet service.)
But within a set of networked mail servers, it could be
a highly effective method of combatting spam.
Note: This may idea may run contrary to the concept of
Unix tools--simple utilites that can be chained
together. Nevertheless, it could be a useful tool
however it is implemented.
Logged In: YES
user_id=2788
Can you detail your idea in respect to how sharing should
work? What should it be able to do?
Logged In: YES
user_id=787521
One idea that came to mind is the following.
A trusted list of mail servers could be maintained in a
config file in /etc. These mail servers would communicate
via a specific (or user specified port) via TCP? to
update/synchronize each other's bogofilter databases. I
haven't thought about the exact protocol for exchange of
information, but possibly something similar to the rsync
protocol. Possibly you would want to limit information
exchange to spam related tokens. (A little experimentation
to determine the most efficient and effective method of
exchange would be necessary.)
Assuming each mail server saw slightly different spam, I
think this distributed approach would be quite effective.
In the event of a new install, it would be very easy to
train one machine from a set of other machines. Especially
for an ISP or larger network.
If I can use a metaphor, when one machine becomes resistant
to a certain spam, it will increase the resistance of the
whole network.
Here is one scenario I can envision, based on this idea:
bogofilter -p /etc/bogofilter.peers [other options]
Periodically, bogofilter would connect with hosts in the
bogofilter.peers file and send/receive updates.