marfcat submission v. SATE.3 Dovecot 2.0 Beta 6
===============================================

The brief selected report is the first attempt at Dovecot 2.0 Beta 6 for
which there were no CVEs. We used training data obtained from the
Wireshark experiments.

The eport validates with the schemas sate_2010.xsd and
sate_2010.pathcheck.xsd.

The report is:

  report-cweidnoprepreprawfftcos-dovecot-2.0.beta6-test.xml

I've omitted other reports as this is experimental completely on
the unknown data on the currently small knowledge base. I've ran
similar tests on Chrome xxx.50 without training on it and got
a big report with the same configuration that I consider very
noisy and don't include, but can provide if required.

In this version, CWEs were the primary category instead of CVEs.
I've re-run the tests not only on Dovecot and Chrome, but also
back at Wireshark, both versions, with quite satisfactory resutls,
but I am not re-submitting the Wireshark as I think it'd be too
many resubmissions for the same test case.

Log files
---------

All the *.log files are there for references. The volume
of the debug info reduced.

Line numbers problem
--------------------

Line numbers in this submission are still 0 as explained in the
generated <explanation /> -- I am using machine learning and kind
of spectral signatures of vulnerable code which is not parsed or
treated line-by-line but rater line-independent "signal". Unfortunate
effect of that is during processing the line information is lost
(the files are treated as if binary, end of lines, as most of the
blanks are usually filtered away). While I have a methodology
in mind of how to reconstruct the line information, it's not
materialized yet.

Serguei A. Mokhov
mokhov@cse.concordia.ca
