►
From YouTube: Cyber Social Media Security: Exploitation for Intelligence and Social Media Reconnaissance
Description
Date: 06/03/21
Presenter: Marshall Gunter
Institution: Nexalogy
Bio: Marshall is responsible for the engineering teams and architecture that power Nexalogy's advanced artificial intelligence tools. With over a decade of executive engineering leadership experience, Marshall has led engineering teams both large and small to drive and build great ideas into successful products. Marshall's engineering background is rooted in big data analysis. He has personally architected systems for both breadth and depth of data ingestion and throughput, processing billions of data points a day across hundreds of terabytes of footprint. Marshall will ensure the Nexalogy engineering department will continue to deliver the high-quality artificial intelligence products demanded by Nexalogy's clients.
A
Thank
you
for
joining.
In
today
we
have
an
established,
exemplary
speaker,
marshall,
gunter,
who's,
leading
several
technologies
in
exology,
and
I've
had
the
pleasure
to
work
with
zach
devereux
at
nexology
on
several
studies,
looking
at
nato's
military
exercises-
and
I
believe,
two
years
ago,
canadian
federal
elections
and
so
on.
So
it's
it's
a
pleasure
to
have
marshall
to
speak
with
us
before
we
get
started.
I
also
like
to
make
some
announcements.
A
We
have
our
social
computing
annual
conference
coming
up
next
month.
So
if
you
haven't
had
a
chance
to
look
at
the
conference,
I
know
some
of
you
have
in
fact
attended
the
content
in
the
last
years.
Please
feel
free
to
take
a
look
at
the
conference.
I
will
post
the
link
on
our
chat
window
and
we
have
several
interesting
tracks
there.
We
have
some
challenge
tracks
for
disinformation
and
a
specific
partner
data
challenge.
A
So
if
you're
working
on
some
of
the
problems
of
disinformation
code,
19,
disinformation
or
partner
data
sets
then
feel
free
to
take
a
look
at
the
challenge.
The
the
regular
track
papers
have
been
submitted,
reviewed
and
accepted.
So
the
program
and
the
conference
agenda
will
be
made
online
very
soon.
A
If
you
haven't
registered
for
the
conference,
take
a
look
at
the
registration
page.
I
think
we
are
still
within
the
window
of
early
registration,
so
if
you're
interested
in
attending
the
conference
and
looking
at
the
data
challenge
feel
free
to
do
that,
we
have
several
other
interesting
tracks
for
your
students.
So
if
you're
interested
in
sending
your
students,
we
have
a
doctoral
consortium
going
on
and
early
or
late
breaking
papers.
A
So
take
a
look
at
the
conference
website,
and
perhaps
you
will
find
something
interesting
there
to
submit
your
studies
or
emerging
studies
from
your
group
and,
let's
see
if
we
have
anyone
new
that
has
joined
the
call
today
and
haven't
had
a
chance
to
introduce
in
the
past
sessions,
we
can
open
the
floor
for
quick
introductions,
marshall.
I
know
you're
probably
joining
this
first
time
and
I
will
take
the
moment
to
introduce
you,
but
if
there
is
anyone
else
who
hasn't
introduced
themselves
before,
please
feel
free
to
introduce.
B
A
Have
you
maria
the
good
thing
about
the
the
group
is
that
all
the
presentations
are
made
available
online
through
youtube
channel.
So
chris
has
been
tremendous
in
doing
those
things
we
can
share
the
link
if
you
would
like
to
browse
our
previous
or
earlier
presentations
for
all
those
where
we
have
the
speaker
consent.
We
do
record
and
upload
this
on
the
south
data
hub
youtube
channel
and
I
believe
today
we
have
marshall's
consent.
B
A
C
A
A
Marshall
works
at
nexology
and
he's
responsible
for
the
engineering
teams
and
architecture
that
power
nexology's,
advanced,
artificial
intelligence
tools
with
over
a
decade
of
executive
engineering,
leadership,
experience,
marshall
has
led
engineering
teams,
both
large
and
small,
to
drive
and
build
great
ideas
into
successful
products.
Marshall's
engineering
background
is
rooted
in
big
data
analysis.
A
He
has
personally
architected
systems
for
both
breadth
and
depth
of
data
ingestion
and
throughput
processing
billions
with
the
billions
of
data
points
a
day
across
hundreds
of
terabytes
of
footprint
yeah.
We
need
to
take
some
tips
from
you,
marshall.
We
are
struggling
with
all
those
type
of
capabilities,
so
I
will
be
very
keenly
listening
to
to
your
talk.
Marshall
will
ensure
the
nexology
engineering
department
will
continue
to
deliver
the
high
quality
ai
products
demanded
by
nexology's
clients.
So
again,
it's
a
pleasure
to
have
you
marshall,
to
speak
at
the
working
group.
A
C
C
Now,
if
you
want
to
go
up
there
so
today
we're
going
to
focus
on
a
particular
technology
that
nexology
offers
called
nexus
smart
and
what
we
do
with
it,
and
I'm
going
to
give
an
overview
of
the
product
and
some
of
the
use
cases
it's
been
used
for
over
the
years
and
then
I'm
going
to
time
permitting
give
you
a
short
walk-through
of
a
problem
set.
We
worked
on
not
too
long
ago,
so
you
can
see
exactly
how
an
analyst
would
use
the
technology.
C
C
Additionally,
we
work
in
private
in
the
private
sector,
with
companies
like
latte
corporation
and
which
is
a
multi-billion
dollar
conglomerate
out
of
korea.
Samsung
group,
I'm
sure
everybody
knows
samsung
ford
motors,
various
others,
but
today
I'm
going
to
focus
on
the
cyber
security
aspects
of
this.
C
If
you
have
questions
around
what
is
being
done
in
the
private
sector,
with
this
tech
or
some,
you
know
different
applications
of
the
tech-
that's
not
shown
here
in
the
private
sector,
please
feel
free
to
reach
out
to
me
I'll
put
my
I'll
put
my
coordinates
in
the
chat.
So
as
just
so
you
as
a
quick
overview
data,
metrics
and
nexology.
We
have
offices
in
montreal
toronto
and
vancouver
we're
headquartered
we're
incorporated
in
toronto.
Ontario,
canada.
We
also
have
an
office
in
seoul
south
korea.
C
So
let's
get
started
so
this
is
nexus
smart.
So
the
problem
we're
attacking
here
is
cyber
security,
so
there's
more
bots
cyborgs
disinformation
than
ever.
So
how
do
you
locate
and
counter
threats
before
they
get
to
you
and
this?
What
we're
discussing
here
we
work
with
social
media,
any
type
of
mass
media
enterprise
data
such
as
email.
C
So
the
tool
and
the
solution
we
have
is
called
nexus
smart.
It's
an
osint
platform
stands
for
open
source
intelligence,
which
analyzes
massive
amounts
of
data,
to
empower
investigative
professionals
to
focus
just
on
the
trade
craft,
while
automating
away,
you
know
busy
work,
and
we
do
this
through
scientifically
validated
methods,
and
the
idea
is
to
shorten
the
time
to
insight.
So
how
do
you
get
to
actionable
intelligence
faster?
C
Like
I
said,
this
has
been
built
in
conjunction
with
the
canadian
government
defense
research
we've
been
deployed
in
the
field
for
the
past
five
years.
We
build
a
lot
of
this.
On
top
of
you
know
the
latest
ai
methods,
deep
learning,
nlp.
You
know
big
data
analysis,
a
lot
of
statistical
analysis
and
the
idea
is
to
manifest
threats
before
they
get
to
you
know
before
they
become
a
threat.
C
So
here
it
is.
You
take
your
data,
you
apply
these
different
algorithms
in
some
degree,
so
machine
learning,
nlp
entity,
extraction,
trending
analysis,
cluster
classification,
information
extraction,
network
and
graph
analysis
topic
and
text
summarization
and
automated
reporting,
and
that's
what
gives
you
nexus
smart
now.
This
is
what
sets
us
apart
we're
actually
at
this.
We
we're
a
discovery
tool.
C
We
don't
just
do
monitoring
monitoring
is
part
of
the
package,
but
what
we
mean
by
discovery
is
I
mean
donald
rumsfeld
used
to
talk
about
this
in
the
bush
administration
and
he
would
talk
about
turning
unknown
unknowns
into
known
knowns.
That's
what
we
do.
We
can
show
you
what's
going
to
bite
you,
you
know
down
the
road.
What's
coming
in
two
days,
what's
going
to
hit
you
in
a
week
what
you
need
to
be
looking
at
now
in
order
to
be
ready
for
then.
C
C
some
of
these
examples.
I
can't
show
you
everything
because
it's
classified,
but
I
will
show
you
what
we're
allowed
to
and
stop
me
if
you
have
any
questions,
try
to
juncture.
So
this
was
the
largest
nato
joint
forces
operation
since
the
cold
war.
There
were
years
of
planning
that
went
into
this
and
it
was
set
to
take
place.
You
know
in
the
north
atlantic,
in
the
north
sea
and
through
the
scandinavian
countries
and
a
year
into
logistics,
for
this
nato
started
to
get
cold
feet.
C
Basically,
nato
wanted
to
ensure
that
they
didn't
get
a
lot
of
info
disinformation
in
there
from
the
russian
counterparts
I.e
when
they
moved
in
thousands
and
thousands
of
tanks
and
planes
and
soldiers
and
boats
and
everything
else
they
didn't
want
it
to
be
seen,
as
you
know,
a
prelude
to
war,
and
when
nato
spoke
in
the
media
that
their
message
was
getting
out
and
not
some
disinformation
network,
and
so
we
went
out
and
we
started
helping
them
with
this
six
months
before
the
operation
started.
I
it
was
a
success.
C
What
you
can
see
here,
whoops
we'll
get
to
it
in
just
a
second.
So
we
start
off
with
like
lexical
analysis.
This
is
some
proprietary
algorithms.
We
use
to
start
surfacing
unknown
unknowns
and
other
things
that
are
trending
in
media
and
social
media.
So
this
is
at
the
beginning
of
the
exercise.
You
can
see
the
clusters
start,
bringing
when
we
start
clustering,
social
media
posts,
or
you
know,
media
articles
and
stuff.
We
can
tell
what's
gonna
start
surfacing
shortly,
and
this
is
one
of
the
tools
we
do
it.
C
You
can
see
that
the
the
computer
and
the
algorithms
color
code
different
things
that
the
analysts
should
be
paying
attention
to
like
this
is
getting
ready
to
start
trending.
You
should
be
looking
at
it
we'll
see
some
more
of
this
later,
but
we
were
successful
with
this,
as
you
can
see
by
the
actor
interaction
graph.
C
This
is
the
social
media
graph
around
the
discussion
of
operation
triton
juncture
and
that
big
green
dot
in
the
middle,
that's
nato,
so
that
when
you
see
a
graph
in
this
shape
it
re,
it
means
that
nato
is
act
really
controlling
the
message
here.
So
when
they
speak,
their
message
is
going
out:
it's
not
being
disjointed
and
fractured
by
disinformation
networks.
There
was
a
bit
of
that
that
you
can
see
in
the
lower
right,
but
not
much
at
all,
so
they
were
successful
by
and
large
in
controlling
their
message.
C
So
when
nato
they
became
the
most
relevant
influencer
inside
the
discussion.
Now
everything
went
completely
well
up
until
the
very
last
day
and
we'll
get
to
that
in
a
minute.
So
that
leads
us
to
another
thing
that
we
do
crisis
management
so
think
black
swan
events
stuff.
You
can't
predict
so
at
the
very
last
minute.
On
the
last
day,
the
helg
inkstad,
which
is
a
norwegian
frigate,
ran
into
a
greek
oil
tanker,
and
so
this
sent
all
the
disinformation
networks
flying
into
operation.
As
you
can
see
on
the
left,
I
mean
sputnik.
C
You
know
religious
affiliations,
works
of
art
stuff
like
that,
and
then
the
computer
can
also
produce
a
summary
of
what's
going
on
so
that
an
analyst
you
know
doesn't
have
the
time
to
read
the
hundreds
of
thousands
of
posts,
or
you
know,
media
documents
that
are
going
on
at
the
time
they
can
rely
on.
You
know
our
tooling
to
get
a
handle
on
the
situation
and
get
this
type
of
information
into
the
hands
of
decision
makers
in
minutes
and
not
hours.
C
So
we're
going
to
see
that
here
this
is
part
of
the
summarization
engine
on
the
right.
This
was
an
early
beta
of
it.
This
is
now
in
full-fledged
it
full-fledged
in
the
product,
but
one
of
the
most
interesting
things
that
came
out
of
this
is
the
computer
picked
up
on
this
before
anybody
else
did
any
intelligence
network.
Nobody,
it
came
out
of
here,
first
and
you'll,
see
we've
highlighted
it
in
the
lower
right.
C
The
the
ai
was
saying,
there's
disruptions
to
gps
signals
and
if
you
know
when
we
dug
into
this
the
reason
it
started
telling
us
this
is
because
people
online
were
complaining
that,
like
their
google
maps
quit
working-
and
you
know
they
couldn't
get
a
signal
on
their
tom
tom
and
so
the
computer's
like
there's
gps
destruction.
Well,
it
turns
out
the
russians
were
actually
just
jamming
one
gps
band
and
it
also
resulted
in
a
whole
bunch
of
planes
having
to
land.
And
you
know
they
never
said.
This
was
what
resulted
in
that
shipwreck.
C
Additionally,
we
do
like
event
monitoring.
So
if
you
need
help
with
your
your
event
or
you
know
something
physically
happening
in
the
real
world,
this
was
a
canada
day.
This
is
the
150th
anniversary
of
canada
and
that's
parliament
hill.
You
see
there
for
those
who've
never
been
here
in
ottawa.
It
was
a
big
deal.
You
know
the
prime
minister
was
going
to
speak
u2,
you
know,
bono,
the
band
and
everything
was
going
to
play
on
that
stage,
and
the
problem
was
the
rcmp,
which
is
canada's
federal
police
force.
C
C
You
know
for
threat
vectors
coming
in
on
social
media
and
they
were
able
to
successfully
do
this
with
our
tooling
with
just
two
officers,
and
this
was
a
big
deal,
because
just
30
days
before
this
one
of
the
guards
on
a
capitol
hill
got
stabbed
and
killed.
So
they
were
really.
You
know
really
serious
about
security.
C
C
So,
of
course
you
know,
intelligence
gets
involved
and
what
happened
was
this
tweet
came
out
and
this
is
on
a
verified
saudi
account
and
it's
basically
a
911
style
attack
warning
against
toronto.
That's
the
cn
tower
in
toronto,
for
those
of
you,
unfamiliar
with
it
and
you've
obviously
got
a
plane
getting
ready
to
fly
into
it,
and
you
can
see
the
arabic
saying
there.
He
who
interferes
with
what
doesn't
concern
him
finds.
What
doesn't
please
him
so
they
said
that
they
came
to
us
and
said:
okay
get
on
it.
So
this
brings.
C
You
know
we
brought
out
one
of
our
other
pieces.
We
do
propaganda
filtering,
so
we
can
filter
out
propaganda
or
focus
on
the
propaganda
in
the
in
our
propaganda
detectors.
So
we
started
on
this
and
one
of
the
interesting
things
we
found
and
here's
an
example
of
predictive
ocean
and
again
of
surfacing
things
the
unknown
unknowns.
C
C
C
This
is
a
five
eyes
event.
The
five
eyes
nations
are
britain,
u.s,
canada,
australia
and
new
zealand.
It's
an
intelligence
sharing
agreement
but
q.
What
they
do
is
they
go
in
and
they
drop
troops
into
contested
environments
and
drones
and
all
kinds
of
other
stuff
and
run
them
through,
and
then
they
monitor
the
populace
to
see
how
they
react
and
they
use
this
to
produce
battle
plans
for
when
they
actually
have
to
do
it.
So
you
can
see
the
soldiers
running
around
that
long
thing
sticking
out
under
his
arm.
C
It
looks
like
some
sort
of
rifle
when
you
see
it,
but
it's
actually
a
camera,
so
they
can
take
pictures
while
they're
down
there.
That's
us
in
the
lower
right.
He
talked
about
zach
zach's,
the
guy
in
the
red
shirt
right
there,
along
with
some
intelligence
analysts
there
so
yeah
we
can
monitor
for
baselines
and
see.
If
something,
you
know,
goes
out
of
baseline
and
you
should
be
looking
into
it
as
far
as
events
on
the
ground
and
situational
awareness
are
concerned,
we've
done
force
protection,
so
we've
monitored
here.
C
This
is
us
being
deployed
in
menusma,
which
was
a
operation
in
mali,
led
by
the
u.n,
and
we
were
there
to
monitor
for
troops
on
the
ground
while
they
were
moving
in
again
a
lot
more
situational
awareness.
What
are
people
doing
as
the
troops
are
moving
in
on
the
ground?
How
are
the
populists
reacting?
What
are
they
saying?
C
You
can
see
here
that
we
uncovered
somebody
some
guys
called
muslim
sniper.
These
were
investigated.
Yes,
so
yeah.
We
can
go
online
and
track
radicalization
as
it
spreads
or
any
other
message
for
that
matter,
as
it
spreads
and
we'll
see
some
of
that
later,
when
we're
combating
disinformation
and
fake
news.
But
it's
the
same
thing:
if
you
can
track
one
piece
of
news
and
how
it
morphs
you
can
track
another.
So
we
do
a
lot
of
you
know,
radicalization
targeting
online
and
seeing
how
terrorism
spreads
around
the
globe
through
online
social
media
efforts.
C
C
C
Well,
the
previous
and
the
current
us
administrations
around
covet,
19
disinformation
again,
there's
another
published
paper
on
this
one
on
the
website.
You
can
see
that
was
co-branded
with
I
forg
forget,
which
u.s
agencies
co-branded
this
paper,
but
you
can
see
it
in
there,
but
it
is
talking
and
mapping
chinese
disinformation
about
the
origins
of
the
covet
19
pandemic
and
again
see
the
paper
some
some
of
it's
redacted
again,
but
it's
interesting
especially
for
professionals,
and
so
we
tracked
a
lot
of
hate
speech
and
we're
able
to
track
these
down.
C
So
here
it
is
nexus
smart
and
the
one
of
the
more
interesting
things
about
this
in
our
tech
is
you
can
automate
all
this
so
instead
of
having
to
go
back
and
look
at
things
all
the
time
you
can
set
up
an
automated
report,
you
can
set
up
triggers
and
you
can
tell
it.
You
know,
report
every
so
often
or
you
can
tell
it
to
report
on
your
triggers.
C
You
know,
updates
of
you,
know
certain
words
or
increase
in
activity,
there's
all
kinds
of
different
triggers
you
can
think
of,
and
these
reports
can
be
automated,
and
this
is
some
of
you
asked
about
new
methods.
This
is
a
it's,
not
an
analysis
method,
but
this
is
a
huge
time
saver
for
everybody
who
uses
the
tooling
now
is
they
don't
have
to
go
back
and
look
every
now
and
then
they
can
just
get
this
sent
to
their
email
or
to
their
smartphone?
C
And
you
know
it
pushes
the
information
to
you.
You
don't
have
to
pull
it
out.
So
that's
what
smart
stands
for
social
media,
automated
reporting
technology
and
that's
it
nexus
security.
Our
aim
is
to
aid
any
intel
team
and
shorten
the
amount
of
time
and
human
resources
it
takes
to
garner,
actionable
intelligence
and
turn
that
into
insights
that
you
know,
then
decision
makers
can
be
used.
Can
use
to.
You
know,
make
more
informed
decisions.
C
C
C
Okay,
can
we
see
this
new
one
big
and
purple?
Yes,
good?
Okay,
so
this
is
how
it's
actually
done.
Not
just
you
know
the
bullet
points,
so
this
is
a
bit
of
an
overview.
We've
went
over
this,
but
again
you
know
we
reduce
your
time
to
insight.
You
discover
what
you
don't
know.
Identify
disinformation,
highlight
you
or
who
or
what
eliminate
time
wasting.
So
ours
is
a
human
in
the
loop
tool
right.
So
we
don't
eliminate
the
human,
but
we
aid
them
in
getting
to
what
they
need
to
do
faster.
C
C
So
the
scenario
is
this:
that
we
have
an
analyst
that
is
tasked
with
monitoring
social
media
discussions
surrounding
iran.
Israel
conflict
we'll
show
how
the
analyst
tracks
the
conflict
discovers
potential
threats
and
then
can
dive.
You
know
from
the
macro
to
the
micro
and
perform
a
deeper
analysis
of
the
threat
so
step.
One
is
data
collection,
so
we
can't
do
anything
without
data.
So
what
we
do
is
we
identify
keywords.
We
identify
the
sources
we
want,
and
these
are
just
booleans.
C
You
know
any
of
these
words
all
of
these
words
this
type
of
stuff,
and
then
we
identify
the
languages
we
want
to
cover.
You
know,
in
addition
to
those
right
there,
we
also
have
traditional
and
modern
chinese
mandarin
and
japanese,
which
will
be
out
very
shortly.
Some
of
these,
you
may
not
recognize
like
daum
and
naver.
Those
are
those
are
korean
news
sites.
C
And
so
we
eliminate
that
data
scraping,
so
we
pull
in
from
all
the
different
targeted
social
media
and
we
can
pull
in
from
traditional
media.
We
can
even
use
email
or
network
crawlers
any
types
of
stuff
you
want
in
you
know
various
different
languages
and
we
pull
in
all
that
unstructured
text
and
we
start
to
make
sense
of
it
all.
C
A
C
So
this
analyst
is
scheduling
a
report
weekly
to
be
emailed
out
on
a
regular
basis,
and
you
know
in
the
form
of
a
word
doc,
and
so
we
deliver
this
in
word
doc,
pdf
for
jason
for
compatibility.
If
you
want
to
ingest
it
into
another
system,
we're
compatible
with
a
number
of
systems,
that's
a
requirement
when
you
work
with
nato
like
we
work
very
closely
with
the
systems
integrations
at
carnegie
mellon.
C
If
you
guys,
if
you
knew
you
work
with
them,
so
let's
see
okay,
so
here's
an
example
of
how
the
analyst
would
start
accessing
their
reports.
It's
like
they
can
go
in.
They
can
see
their
snapshot
history.
They
can
download
any
of
the
reports.
Previous
reports,
new
reports
and
start
looking
through.
You
know
the
different
analysis
that
the
computer
showed.
C
So
next
data
summarization,
so
this
we
saw
a
bit
earlier
when
the
ai
is
going
through
and
saying,
here's
a
summary
of
what
we
believe
is
going
on,
and
this
allows
for
you
know
a
very
quick
way
for
the
analyst
to
pick
up
and
uncover
potential
problems
right.
So
what
you
see
here?
This
is
a
real
op,
so
this
is
declassified
pieces
of
a
real
op.
So
what
you're
going
to
see
is
they
picked
up
on
this
natanz
and
then
on
the
right?
C
You
know
it
says
you
know:
israel
believes
that
the
natanz
blast
set
uranium
enrichment
capacity
back
15
years,
etc,
and
so
the
the
analysts
wanted
to
start
digging
into
this,
like
you
know
what
what
are
they
saying
online
about
this?
So
next
you
know
once
that
comes
up
in
their
automated
report.
They
say:
okay,
let's
see
what's
going
on
and
dive
deep
down
into
the
issue,
so
they
want
to
access
the
most
relevant
media
posts
and
they
don't
want
to.
C
You
know
the
computer
determines
the
relevancy
and
the
analyst
can
set
parameters
around
what
they
mean
by
relevant,
but
most
of
the
time,
98
of
the
time.
I
would
say
they
just
use
what
we
have
out
of
the
box
and
so
what
they
use
with
this
relevancy
is
they
want
to
see
what
what
is
the
top
shared
content
like
what
is
going
viral
right?
What
are
the
top
words?
What
are
the
top
actors
or
other
relevant
data
slices
used
by
these?
You
know
visualizations,
online
etc.
C
So
here
you
know
you
can
see
israeli
military
officials
united
in
opposition
to
biden's
iran
deal.
So
what
is
that?
Why
is
that
trending
around
natanz?
Why
is
iran
rattled,
as
israel
repeatedly
strikes
key
targets,
so
they
can.
These
are
the
links
that
are
being
shared
online
by
the
most
right
now
that
have
something
to
do
with
this.
C
So
again
the
analyst
is
going
to
go
in
and
you'll
see
here,
they're
going
to
see
the
timeline
they're
going
to
come
down,
here's
some
of
the
shared
links
they
want
to
look
and
you
can
click
on
some
of
these
and
then
you
see
it
goes
to
the
jerusalem
post.
That's
an
example
of
where
that's
shared
here
again
they're
looking
for
the
top
hashtags,
and
so
here
it's
going
to
present
you
with
you
know
the
tweets,
some
of
the
most
relevant
tweets
that
are
going
around.
C
C
And
so
when
we
compared
peaks
and
troughs
you
can
see
that
the
spikes
in
conversations
what's
being
driven
there
and
again
you
can
use
these
graphs
just
like
pre.
We
previously
saw
to
go
from
the
macro
to
the
micro,
and
this
really
reduces
your
time
to
insight
like
it.
This
makes
it
a
lot
quicker
when
you
don't
have
to
go,
hunt
this
stuff
down
and
the
computer
is
pulling
it
out
for
you.
C
So
we've
identified
the
abnormalities
and
then
we're
gonna
go
look
at
some
anomaly
posts
again,
so
we
can
see
here
the
computer
go
in
click
on
it,
so
we
can
they
trick.
I
didn't
see
what
he
take,
but
you
can
pick
most
popular.
You
can
pick
you
know
most
relevant
and
we're
going
to
see
what's
driving
these
spikes-
and
you
know
why
is
this
going
viral
now?
What's
the
type
of
media?
That's
you
know
making
it
do
so
so
again.
C
So
again,
this
is
all
right,
so
the
the
technology
improvements.
This
is
probably
not
relevant
to
this
discussion,
but
I'm
going
to
talk
about
it
just
briefly,
because
you
asked
about
it
earlier
about
developing
new
tech,
these
automated
summaries.
This
was
a
presentation
given
to
various
nato
organizations
not
too
long
ago.
C
C
If
you
go
on-
and
you
just
put
the
words
donald
trump
in
there-
you're
going
to
get
a
swath
of
information,
but
the
idea
is
to
reduce
the
analyst
busy
work
and
eliminate
any
type
of
confirmation
bias
and
that's
really
important.
You
know,
because
the
confirmation
bias
has
been
a
problem
in
the
past
and
by
you
know,
building
these
algorithms
to
do
it.
We
effectively
remove
that
because
we're
removing
the
human
from
that
piece
of
the
loop
and
then
so
topic
summarization
again.
This
is
some
of
the
tool
doing
it.
C
So
you
can
see
there
they're
pulling
on
these
topics
the
election,
and
this
allows
the
analysts
to
go
in
and
see
if
they
want
to
dive
into
the
tool
along
different
timelines
about
why
the
computer
decided
to
or
why
the
algorithm
decided
that
this
was
a
proper
summary
and
give
some
examples
of
you
know
where
it
pulled
information
from
so
it
can.
You
know
you
can
verify
the
computer's
analysis
of
the
of
the
topics
inside
the
tooling.
C
And
that's
it!
That's
a
live
example
of
somebody
using
our
tool
to
tackle
a
certain
situation,
but
again
we
empower
you
to
focus
on
tradecraft,
thus,
shortening
your
time
to
insight,
we
allow
you
to
rapidly
gather
data
from
a
variety
of
sources,
schedule
and
monitor
threats
with
automatic
triggering.
C
We
provide
digestible
insights
to
non-analysts.
This
is
important
as
that
gap
between
how
an
analyst
views
things
and
how
a
non-analyst
views
things
can
be
very
broad,
sometimes
and
we're
always
looking
at
new.
You
know
state-of-the-art
ai
and
machine
learning
techniques
to
make
our
work
and
tools
better
and,
lastly,
we're
again
we're
a
discovery
engine.
A
Yeah,
thank
you,
marshall.
That's
a
fantastic
tool
and,
and
I've
had
a
pleasure
to
work
with
zach,
and
it
is
fascinating.
Every
time
I
hear
about
nexology,
there
is
something
new,
so
that
is
interesting
to
see
how
it
has
advanced
and
grown
tremendously.
Over
time.
We
have
some
questions
on
the
chat.
I
believe
the
first
one
from
shannon
has
been
addressed.
A
C
Yes,
absolutely
so
we
can,
we
can
cut
the
data
almost
any
way
you
can
think
of
that's
one
of
the
strengths
we
have
also
is
the
diversity
of
the
platforms
that
we
can
address.
So
if
you
just
wanted
to
focus
on,
say
facebook,
you
can
do
that.
If
you
want
to
focus
on
twitter,
you
can
do
that
and
then
the
computer
and
the
ai
will
recalculate
everything
based
on
just
those
those
slices
of
data.
It's
not
an
issue.
C
We
can
also
cut
it.
Many
other
ways,
not
just
platforms,
but
we
can
cut
it
by.
You
know
different
types
of
individuals,
right
actors
right
so
do
we
want
actors
in
just
this
one
area
or
do
we
want
actors
of
just
these
origins
right
or
we
can
even
cut
it,
and
this
is
really
interesting
to
the
private
sector
and
they
use
this
a
lot
I'll.
Give
you
a
quick
example:
we
can
actually
cut
it
by
what
people
are
talking
about
like
what
they're
actually
saying-
and
this
is
where
the
private
sector
comes
in.
C
C
So
in
a
normal
marketing
campaign
you
see
about
one
percent,
click-through
rate
we've
seen
click-through
rates
up
to
17
percent
on
these.
So
it's
very
effective.
It's
also
used
by
the
rcmp
for
this.
They
call
it
the
be
on
the
lookout
program.
They'll
use
it
to
target
for
criminals
they're
looking
for
in
an
area,
so
they'll
find
all
the
people
that
are
talking
about
crime
and
criminals
and
then
they'll
send
out
a
bolo
to
them
and
say:
have
you
seen
this
person
and
those
people
are
more
likely
to
share
it
in
their
circles?
C
A
B
C
Motivation
is
a
hard
piece
to
discover,
like
you
know
what
hides
in
the
hearts
of
men,
I
can
say
from
the
work
we've
done
at
a
state
level,
so,
like
an
actor
as
a
state,
you
can
see
the
results
of
what's
being
had
and
you
the
motivation,
is
to
fracture
and
divide.
Now.
Let
me
give
you
some
examples.
We've
been
involved
in
obviously
the
election,
the
recent
presidential
elections
in
the
u.s,
definitive
state
actor
interference,
definitive.
C
You
know
the
whole,
I
mean
it
was
very
effective
all
throughout
the
2010s
and
if
you
saw
in
2014
before
people
were
actually
talking
about
this,
it
was
already
happening
because
you
know
the
russian
state
was
already
serving
up
information.
We
track
it
back,
as
the
russians
been
doing
this
since
the
50s,
but
we
can
track
it
back
to
social
media
all
the
way
back
to
like
about
2009
2008.
C
Their
motivation
is
to
divide
it's
to
keep
populations
divided
and
to
keep
them
from
focusing
on
other
goals.
Russia
was
able
to
successfully
annex
crimea
and
nobody
said
boo
because
everybody
was
too
busy
with
their
infighting
and
it's
to
exacerbate
those
problems.
They
don't
really
care
about.
I
don't
know
to
give
some
u.s
examples.
C
They
don't
care
about
black
lives
matter
right
they
just
don't
they
don't
care
about
republicans
or
democrats
or
any
of
this
other
stuff.
What
they
care
about
is
these
people
are
at
each
other's
throat,
and
you
know
we
saw
this
all
over
the
place
like
we.
There
was
the
largest
group
online
that
was
advocating
for
texas
to
secede
from
the
united
states,
turned
out
to
be
a
russian
led
op,
so
it's
divide
and
conquer.
B
C
C
Started
you
know
on
both
sides
of
the
fence
they
started
stoking.
You
know
the
albertans.
They
started
stoking
the
quebecois,
which
are
the
people
of
quebec,
and
it
was
just
nasty
for
like
three
years
and
it's
it's
the
results
linger
today
I
mean
we
still
hear
people
on
the
street
talk
about
how
how
horrible
alberta
is
and
people
in
alberta
talking
about
how
horrible
the
quebecers
are,
but
I
mean
we're
all
the
same
country.
So
at
a
state
level,
you
know
they're
working
on
state
goals
on
an
individual
level.
I
mean
to
be
brutally
honest.
B
C
Yeah,
it's
absolutely
dead,
it's
absolutely
able
to
trace
it
back
to
specific
posts
and
specific
individuals,
and
you
saw
some
of
that
during
the
demo
where
we
went
from
you
know.
I
talked
about
going
from
the
macro
to
the
micro,
so
if
you
want
to
see
who's
doing
it-
and
we
have
you
know
graphs-
you
saw
in
the
the
in
the
trident
juncture.
I
I
put
up
one
the
one
representation
of
the
social
media
graph,
all
those
little
blue,
dots
and
stuff.
B
C
C
B
C
That
up
that's
what
they
call
them,
and
these
are
the
people
that
just
regurgitate
whatever
comes
online
or
they
reframe
it
and
they
have
built
into
the
messaging
just
just
enough
to
where
people
actually
reframe
reframe
it
in
terms
of
their
politics
or
their
religion
or
whatever
point
they're
trying
to
make
right
that
it
becomes
even
more
difficult
to
track
that
narrative.
But
it
still
has
the
same
effect.
B
C
C
C
Well,
we
can,
we
can
assist
in
that
now
it's
not
automated
or
anything
like
that,
but
the
platform,
a
lot
of
the
tooling,
is
built
around
a
certain
protocol
called
bend
like
b-e-n-d,
and
this
is
the
narrative
that's
in
the
protocol,
that's
in
use
by
nato.
It
stands
for
build
and
gauge
neutralize
distract.
C
Now
we
do
sit
on
a
nato
working
group
with
participants
from
22
other
countries
around
disinformation
and
combat
combating
it,
but
nato
has
determined
and
rightfully
so
that
you
know
once
something
gets
online.
You
know
you
can't
do
anything
about
it.
You
know,
10
years
ago
people
were
trying
to
scrub
stuff
off
the
internet.
You
just
can't
do
it
once
it's
out
there,
it's
out
there,
especially
with
the
the
penetration
we
all
have
now,
where
everybody
everybody
who's
got.
C
A
smartphone
is
now
a
publisher
of
information
right
right,
so
everybody
has
a
voice
and
a
platform.
So
the
idea
behind
it
is
well.
What
can
we
do
to
build
and
engage
our
narrative
and
neutralize
and
distract
from
their
narrative
right?
And
so
how
do
you?
How
do
you
get
your
message
to
be
the
one
people
are
listening
to
and
not
the
other
guy.
B
C
So
yeah
we
have
built
a
lot
of
our
tooling
around
that
protocol,
and
so
we
can
be
a
part
of
the
puzzle
right
we're
not
the
complete
puzzle.
You
know
puzzle
but
we're
a
part
of
it
and
we'll
allow
analysts
and
you
know
other
people-
you
know
the
tooling
if
they
want
to
engage
in
that
type
of
that
type
of
behavior
they
can,
and
it's
also
in
a
well
like
I
said
in
the
private
sector.
C
It's
a
very
effective
protocol
for
ad
marketing
turns
out
that's
what
samsung
you'll
see
them
on
our
list
of
clients.
They
do
a
lot
of
ad
marketing
and
they
do
they
monitor
their
employees,
all
their
employees
email
through
our
system
and
they
use
it
to.
You
know
figure
out
if
one
particular
group
is
having
a
hard
time
or
one
group's
doing
well
things
like
that.
Oh.
B
B
C
C
C
A
Thank
you,
marshall.
I
had
a
quick
question
and
perhaps
if
there
is
time
I
can
ask
my
second
question
also
in
the
studies
that
we
have
conducted
for
canadian
election
and
I
believe
zach
was
also
looking
in
some
of
the
work
that
we
had
done
on
youtube
and
disinformation
campaigns
during
canadian
2019
prime
ministerial
elections.
A
C
C
Well,
we've
run
into
them
at
a
corporate
level.
Okay,
they've
never
gone
after
our
staff
directly,
but
the
same
group
that
you're
talking
about
is
taking
pot
shots
at
us
because
if
we
ever
name
them
in
a
piece-
and
we
have
several
times
they
they
get
bent
out
of
shape.
So
I
guess
this
would
be
the
equivalent
in
the
u.s
of
what's
that
guy's
name.
C
Yeah
yeah
they're
very
similar
in
the
way
they
operate,
but
yeah.
So
no
we've
never
had
a
problem
with
it.
I
think
one
of
the
reasons
we
don't
have
a
problem
with.
It
is
a
lot
of
times
when
we're
naming
this
stuff.
I
mean
it's
co-branded
by
say
you
know
defense,
research
or
the
rcmp
and
if
I
don't
think
they
want
to
take
those
entities
on
quite
honestly.
A
Maybe
that
their
operations
are
evolving
because
for
our
city
that
was
published
by
royal
canadian
forces
early
or
late
last
year,
what
they
did
was
they
basically
started
harassing
us
with
foia
requests
first
and
then
started
harassing
office
of
naval
research,
national
science
foundation
and
other
funding
agencies
who
have
supported
that
work
through
foia
requests.
But
of
course
that
is
canadian
group,
so
they
cannot
technically
do
a
foia
request
for
a
u.s
entity,
so
they
hired
local
little
rock
based
companies.
That's
where
I
am
so.
A
They
hired
a
little
lock-based
companies
to
do
those
foia
requests
and
then
they
didn't
stop
it
there,
because
we
answered
those
requests.
They
started,
sending
out
harassing
emails
to
me
and
some
of
the
other
program
officers
who
were
acknowledged
on
the
brand,
so
technology
on
the
paper.
So
that's
that's
what
I
was
wondering
so
you
haven't
had
such.
C
C
Yeah,
I
just
I
honestly,
I
personally
ignore
it
the
as
far
as
the
company
is
concerned.
Yes,
we've
been
harassed,
we've
we've
actually
been
named
and
you
know
they've
published
articles
trying
to
disparage
our
work.
You
know,
and
you
know,
to
mar
our
name
and
those
articles
have
come
out
both
in
the
u.s
canada.
C
So
we
rely
on
that
and
to
make
sure
that
you
know
we're
in
the
right
before
we
go
ahead,
see
what
I
mean
yep
yeah.
So
that's
that's
one
of
the
ways,
probably
the
primary
way
that
we
defend
ourselves
against
anything
like
that.
That
comes
up
is
we're
not
talking
about
opinions
here,
we're
talking
about
facts
and
truth,
and
here
it
is.
A
There's
no
point
arguing
with
them,
because
they're
not
going
to
debate
on
the
argument
or
the
rational
or
the
logical
facts,
so
yeah
they've,
you
know
gone
on
to
the
extent
that
they
had
started.
Sending
emails
to
my
colleagues
sen,
saying
that
the
research
that
comes
out
of
this
group
is
biased,
and
this
is
not
fair,
so
they
try
to
disparage
your
name
or
the
work
that
you're
doing,
even
though
this
was
a
highly
reviewed
publication
and
published
by
royal
canadian
forces.
A
Yeah,
well,
I
know
that
we
have
short
our
regular
time,
so
I'm
sure
we
have
folks
dropping
out
and
thank
you
again,
marshall,
for
this
fantastic
talk
and
you
have
shared
the
link,
so
that
link
will
also
be
shared
with
all
the
working
group
participants
and
yeah.
Thank
you.
You,
thanks
to
you
and
to
the
participants
who
joined
the
calm,
and
I
look
forward
to
seeing
you
next
month.