[GRLUG] $50K Prize if You Find Way to Block Robocalls

[Michael Mol]

On Wed, Nov 7, 2012 at 11:44 AM, Tim Schmidt <timschmidt at gmail.com> wrote:
> I think the answer is doubly that they have more data, and that they
> keep it secret.
>
> Sharing the data means sharing it with the spammers, and in my short
> time as an email admin, subscribing to the popular spam lists only
> blocked about half our incoming spam.  Even with greylisting, lots of
> filter tweaking for company-specific domain terms, scads of training,
> and measures I've since forgotten, we still regularly received spam.
> Those guys are really good at what they do.

Yes, having the data secret makes it more difficult for spammers to
know how to work around it.

At the same time, having the data shared makes it easier to accumulate
a corpus that's more difficult to work around. Having additional types
of data that are shared increases that level of difficulty
exponentially.

For example, consider sharing keyword lists known to primarily exist
in spam data. Consider sharing keyword composition analyses showing
codepoint makeup for messages which are typically spam. Consider
showing header chains which are typically spam.

The more dimensions you come from, the more likely their text munging
algorithms will cross a previously-safe boundary while they try to fit
within a new one.

--
:wq