MT an open relay
It seems MT default install acts as an open relay. Every spammer’s dream: kasia in a nutshell: MT an open relay
Ham, Spam … and more
Via mac.against.org comes something that I already asked for some time ago. A bayesian filter capable of separating not only ham from spam but also filter mails into their correct folders, learning just by example. Too bad Outclass it’s Windows/Outlook based.
mac.against.org - Windows/Applications/Outclass Outclass is a Bayesian classifier for Outlook (it relies on POPFile to do most of the actual work, but works seamlessly with Outlook in an Exchange-only environment). Besides the usual Spam/non-Spam classification, it can also classify e-mail into several “buckets” or categories, and with very little training it soon began to classify my e-mail per project, with no direct mention of the project name or type in the messages - it simply inferred it from its previous training, and without explicitly associating the senders with the project (which you can also do).
Japanese Spam
There always a first time for everything. Even for receiving
japanese (is it?) spam (or whatever that is).
src="http://www.andrerestivo.com/weblog/archives/japanspam.png"
width="377" height="276" border="0" />
Spam, Lovely Spam
href="http://www.americanscientist.org/Issues/Comsci03/03-05Hayes.html">
Spam, Lovely Spam: Seems to be a nice article about the Spam
issue. Have to find time to read it tomorrow. Must sleep now …
Zero Spam
Only 7 days ago I
href="http://www.andrerestivo.com/weblog/archives/001912.html#001912">
moved from Evolution to Mozilla Mail. At that time I said that
I was looking forward to the day I would have zero spam mails in my
Inbox. Well, that day arrived sooner than I expected. And besides
having no spam I had no false negatives either. A big kudos for
theMozilla Spam
Filtering team.
Still there are some stuff that could be improved:
- Use bayesian filters not only to differentiate ham from spam
but also has a substitute for filters - Train those bayesian filters automatically when I move a
message from my inbox to a folder
That would be ultra-cool as it would be very easy for Mozilla to
separate all my mailing lists and newsletters without me creating
filters by hand.
Spam Inflection Point
href="http://www.tbray.org/ongoing/When/200x/2003/03/03/Spam">Spam
Inflection Point?: Are we finally winning the war against spam?
I just moved from Evolution to the new Mozilla 1.3 and I’m looking
forward to the time I’ll have 0 (zero) spam in my Inbox.
TarProxy
href="http://www.martiansoftware.com/articles/spammerpain.html">TarProxy:
A fresh idea in the fight against spam.
Now we have moved identification of spam to the time of
its receipt. But how can the SMTP server best use this knowledge? I
propose that the running probability from the classifier be used to
throttle the connection with the offending server. If an incoming
message looks like spam [1], the connection could be slowed
dramatically, consuming the spammer’s resources and wasting their
time [2]. This would transform the server into a sort of dynamic
tarpit, in which the spamminess of the incoming message affects the
viscosity of the tar [3]. As the spam probability goes up, the
socket speed goes down [4].
Microsoft goes after spammers
href="http://www.theregister.co.uk/content/6/29382.html">The
Register: Finally Microsoft does something usefull.
Anti spam idea
Some time ago Sergio
his requirements for the perfect anti-spam system. One of them was
this:
Zero false positives - Losing one
valid message is always worse than one spam message passing
through. Several orders of magnitude worse. It’s possibly
catastrophic. It is a risk that can’t be taken. It must be
matematically provable that the system won’t reject a valid
message.
As I pointed
this requirement is the major fault of AI based spam filtering. A
single false positive could destroy a good spam filter completely
(see
article for a recent example).
All this to say that I had a crazy idea. Maybe it has already
been discussed and probably has lots of problems I haven’t thought
about. But here it goes:
Why not have the anti-spam filter reply to the message sender
with a message explainig that his message has been filtered and
requesting that he resends the message to a specific, one time
only, e-mail address? Spammers would get all their e-mail back and
someone trying to contact you would have a way of sending you an
e-mail even if it gets filtered.
An example:
- jonh_doe@example.com sends an e-mail to
joane.doe@example.com - For some reason Joane’s e-mail reader rejects the e-mail as
spam and sends a reply to John telling him to resend the message to
joanne.doe-617131243@example.com - John does has told and the message his delivered
- The generated e-mail address is deleted
This solution combines the best of spam filters and systems
requiring the user to prove he is human (like
Arrest). Could this work?
A plan for Spam
A nice spam filtering mechanism explained in this article.
To the recipient, spam is easily recognizable. If you
hired someone to read your mail and discard the spam, they would
have little trouble doing it. How much do we have to do, short of
AI, to automate this process?
From
