[cryptome] Re: diagnose yourselves
- From: coderman <coderman@xxxxxxxxx>
- To: cryptome@xxxxxxxxxxxxx
- Date: Fri, 26 Dec 2014 05:48:53 -0800
On 12/26/14, In Harms Way <11414150173@xxxxxxxxxx> wrote:
>
> No, its not, it just tries to please you.
actually, it's trying to please the marketing types!
this is really just an amped up version of "sentiment analysis", where
"sentiment" is a simplified set of these criteria.
what i find interesting is when these types of systems are applied on
large scale for network attack analysis. "skiddie, n00b, moderate,
skilled" classified attacks, with probable origins.
again, not really insightful, but certainly more informative than your
typical "threat intelligence service" *grin*
in the future, crypters will not just obfuscate the mechanics (for A/V
avoidance), they'll distort the semantics of your attack technique
(for attribution denial).
and of course, what none of these systems tell you is there inherent
limits and severe constraints affecting utility. there have been some
good presentations on "signal jamming machine learning", but i'll have
to dig them up later unless someone beats me to it :)
best regards,
CoderMarkovMan
Other related posts:
