►
From YouTube: ROS 2 Security Working (12 Apr 2022)
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
A
Yeah,
well,
it's
they're
very
strong.
It's
part
of
the
way
it
wouldn't
be.
So
we
we
thought
to
organize
this
meeting
as
a
study
group,
the
discussion
group
or
on
the
paper
itself,
but
also
everything
it
implies.
A
What
what
comes
next
I
guess,
but
I
lived
before
to
referencia
Francia.
If
you
should,
like
she's,
been
studying
closely
in
the
paper
and
she's
prepared
a
few
direction
for
the
for
the
discussion.
B
Yeah
so
again
like
congratulations
for
putting
together
this
this
paper.
As
we
all
know,
it
goes
in
the
same
direction
as
the
group
has
been
pushing
to
to
promote
disability
and
Adoption
of
astros2.
So
that's
pretty
great
and
very
well
aligned
with
our
work
here.
So
basically,
the
proposal
yeah
was
to
have
a
discussion
around
the
paper.
B
B
Questions
in
the
agenda
just
to
guide
discussion
doesn't
have
to
be
limited
to
this,
but
basically
taking
in
the
direction
of
how
can
this
work
help
promote
usability
and
Adoption
of
festivals
too,
for
example,
as
a
first
question,
so
I
basically
throw
in
this
this
reflection,
question
and
I
guess
maybe
you
have
some
insights
or
some
ideas
on
how
you
plan
to
to
make
this
happen.
B
So
whoever
wants
to
jump
in
and
ensure
some
ideas
on
this?
It's
very
welcome.
D
C
Roughing
it
sounded
like
you
wanted
to
okay
I'll,
just
throw
a
few
thoughts
so
well.
C
First
of
all,
I
wanted
to
actually
say
that
this
is
awesome,
so
thank
you,
Florencia
for
taking
the
lead
and
also
thank
you
Jeremy
for
empowering
all
of
us
with
this
I
think
we
should
definitely
have
more
light
study
sessions
like
this,
and
if,
if
this
is
kind
of
like
an
open
topic,
that's
of
interest,
then
I
can
suggest
a
number
of
papers
that
I've
been
collecting
and
reviewing
and
a
number
of
authors.
C
C
Security
discussions
which
are,
in
my
opinion,
are
very
much
of
interest
to
to
these
working
groups,
activity
so
yeah
and
regarding
the
first
cool
all
right.
So
let's
discuss
that,
maybe
in
the
chat
or
on
the
side,
happy
to
help
with
that.
Regarding
the
first
question
raised
in
here
and
by
the
way
I
like
the
questions,
very
much
I
especially
have
a
few
thoughts
of
interest
on
the
last
one
and
I.
Actually,
I
am
glad
that
this
came
up
and
I
think
this
can
trigger
corporations.
C
So
I
will
wait
until
we
reach
there
to
drop
a
few
ideas,
but
regarding
the
first
one
well
in
the
paper,
essentially,
as
you
can
see
in
the
title,
we
pretty
much
focus
on
on
usability
one
of
the
biggest
concerns
we
have
received
over
quite
a
few
years
now,
or
at
least
that
I
have
received,
while
speaking
with
industry
using
Ross
and
considering
to
use
Ross,
it's
essentially
that
it
came,
it
came
to
quite
a
challenge
to
start
using
Ross
in
a
secure
manner,
specifically
because
the
guidelines
or
the
lack
of
guidelines
first,
that
came
as
a
constant
concern.
C
Second,
because
the
documentation
wasn't
as
maintained
as
probably
some
customers
or
users
may
want
it
to
be-
and
this
is
certainly
our
fault,
but
also
it
came
down
to
the
fact
that
after
a
number
of
years
working
on,
thank
you
on
on
this
topic
itself.
C
We
realized
that
actually
most
users
weren't
quite
addressing
the
major
concerns
from
a
security
perspective,
not
so
much
because
they
didn't
want
it,
but
more
because
essentially
they
they
didn't
have
the
facilities
to
do
so
and
that's
how
it
Dives
down
into
usability.
This
is
articulated
nicely
into
the
introduction
of
the
actual
article.
You
know
the
actual
paper
and
it
all
boils
down
to
actual
reason,
if
you
think
about
cyber
security
and
that's
part
of
what
I'm
doing
in
my
day
job
actually
these
days.
C
C
That
means
that
the
technical
complexity
of
these
actual
attacks
in
the
real
industry
out
there
don't
really
involve
much
technical
complexity
and
it's
just
a
matter
of
training,
conveying
the
right
information
about
security
practices
across
the
whole
supply
chain
and
particularly
robots
as
a
major
Target
increasingly
so
that
motivated
us
very
much
and
also
I,
think
motivates
the
straightforward
answer
to
to
to
this
question
itself,
which
is
how
can
we
help
promote
more
usability
of
srs2
and
I?
C
Think
one
of
the
answers
we
came
up
while
doing
this
work
was
by
trying
to
propose
a
methodology
methodology
that
helps
steer
security
processes
from
security,
cyber
security
Engineers
all
the
way
to
roboticists,
while
confronting
robotic
setups.
So
that
is
the
first
that
comes
to
mind
at
least
to
me,
but
I'm
sure,
Jen,
Luke
and
Rafi
may
have
more
comments.
C
A
second
that
I
would
say
is
that,
while
doing
this,
actually
we
identified
lots
of
interesting
opportunities
that
maybe
we
just
didn't
reflect
enough
to
consider
in
that
writing
them
down
actually
helped
significantly.
C
This
is
written,
particularly
in
the
paper
in
the
fifth
section
in
the
conclusion,
and
we
really
really
had
like
a
decent
I
would
say:
brainstorming
on
the
side
of
this
group,
where
we
through
some
really
nice
ideas
on
what
we
think
might
be
interesting
to
improve
the
overall
usability
and
usage
of
the
tooling
of
the
tools
that
S
Voice
is
all
about,
and
some
examples
include,
for
example,
things
such
as
including
more
capabilities
to
introspect
computational
graphs.
C
This
is
at
the
bottom
and
at
the
core
of
the
security
processes
themselves,
things
such
as
maybe
extending
some
of
these
tools,
not
necessarily
due
to
DDS
itself,
but
also
to
embrace
other
communication
middlewers
we're
increasingly
seeing
how
some
companies
are
actually
replacing
the
middleware
that
leaves
below
the
Ross
Middle
over
layer,
rmw
and
how
this
is
going
to
get
more
and
more
important.
C
As
as
again,
we
confront
more
and
more
difficult
use
cases
involving
potentially
many
more
attack,
vectors
and
entry
points
for
attackers,
and
the
last
thing
I
would
mention
is
which
I
think
it's.
It's
definitely
something
that
rafin
and
Jen
Luca
have
been
thinking
a
lot
about
so
I
credit
them
back
on
this.
But
it's
the
idea
of
this
graphical
user
interfaces,
which
I
think
syncs
very
nicely
to
some
of
the.
C
In
a
way
complaints
I
think
we
received
in
this
working
group
in
past
meetings,
I'll
just
stay
at
that
and
I'll,
let
Jen
Lucan
and
Ravin
maybe
add
more
details.
But
hopefully
this
these
ads
are
built
into
into
that
question.
C
E
Thanks
Victor,
so
the
the
two
kind
of
points
I'd
like
to
kind
of
hamper
on
that
maybe
touch
on
the
questions
here
is
some
of
the
blocks
are
the
measurement,
so
that's
sort
of
the
foundation
of
of
how
we
want
to
improve.
E
The
usability
here
is
to
to
try
and
automate
as
much
as
possible,
and
so
we
can
Aid
the
user
in
generating
a
an
accurate
policy
to
try
and
offload
as
much
that
as
possible,
but
that
that's
limited
to
how
accurate
we
can
measure
the
the
computation
graph
and
what
permissions
are
being
used
and
and
how
to
audit.
E
And
so
we
had.
You
know
two
ideas.
You
know
one
was
to
improve
the
Ross
graph
API
so
that
you
know
it
can
give
us
analytics
on
the
more
temporal
interfaces,
such
as
action,
clients
and
such
and
the
other
one
is
logging
and
I
I.
Think
there
was
a
push
a
while
ago
to
to
improve
the
security
logging
another
feature
set.
E
So
if,
if
we
had
either
of
those
two
kind
of
buffed
up,
then
we
could
build
on
that
with
more
tooling
to
to
kind
of
generate
the
on
the
policies
going
forward.
A
It
also
is
like
sorry
Russian
if
I'm
entering
to
just
a
second
before
we
change
the
topic,
I'd
like
to
very
quickly
address
the
two
points
you
you've
raised
already:
the
boats,
mapping
the
the
the
graph
and
and
that's
somewhat
tied
to
the
user
interface
you're
mentioning.
A
This
is
also
the
idea
behind
nodiel
right.
So
nodiel
is
a
declarative.
A
A
declarative
topology
of
your
graph
because
you
know
each
and
every
Ross
communication.
A
In
your
packages,
and
because,
thanks
to
the
launch
file,
you
you
know
beforehand
the
and
when
I
said
before
and
I
mean
right
before
runtime,
you
know
the
the
Rose
craft
at
the
end,
up
to
some
exceptions
with
no
idea
you're
able
to
to
generate
from
the
flight.
The
policy-
and
this
is
already
working
for
very
simple
examples.
We
are
looking
at
putting
together
a
much
larger
example,
making
use
of
mobile
more
complex
and
competing
example.
A
D
A
They
are
using
on
the
navigation
working
group,
so
they
have
the
so-called
growth
software,
which
is
a
nice
graphical
interface
to
put
together
your
behavioral
tree
and
I
do
believe
that
we
can
reuse
at
least
some
of
these
software.
The
group
software,
but
instead
of
putting
together
well
behaviors.
B
A
A
E
So
the
the
node
ADL
and
cautiously
optimistic,
the
when
I
was
doing
some
some
more
recent
experiments
with
either
the
navigation
stack
or
or
the
move
it
stack.
I
was
trying
to
build
an
entire
policy
there,
it's
pretty
challenging
and
that
a
lot
of
the
developers
are
now
migrating
that
there
was
a
lull
where
there
wasn't
any
XML
based
launch
file.
Then
the
static
launch
files.
E
So
then
you
know
python
became
sort
of
the
de
facto
and
that's
like
a
fully
turned
complete
language
and
so
having
to
interpret
that
to
figure
out
I
I
I'll
have
to
check
the
the
current
status
on
how
your,
how
you're
extrapolating
what
what
interfaces
are
being
used
purely
from
the
launch
files,
but
it
does
seem
pretty
pretty
ambitious,
and
so
one
thing
that
might
be
also
good
as
a
as
a
sanity
check
is
to
still
have
the
measurement
tooling
to
kind
of
validate
your
node,
your
your
node
DL
as
well,
so
make
sure
your
node.l
doesn't
miss
anything
or
it's
out
of
sync.
E
With
respect
to
the
current
code
base
and
talking
about
current
sinks.
E
That's
something
I
also
noticed:
John
Luca
has
a
has
a
master's
student
that
sort
of
tasked
to
kind
of
Implement
sros
on
you
know
one
of
their
own
projects
and
one
of
the
things
they
encountered
was
a
pretty
large
desync
between
the
current
documentation
and
current
implementation
and
I
think
it's
just
sort
of
an
inevitable
symptom
of
you
know:
open
source
development
and
acceleration
and
how
fast
we
move,
but
I
think
it's
also
a
a
case
example
for
motivating
certain
moving
users
away
from
having
to
know
the
Nitty
Gritty
details
of
how
the
policy
documents
are
formulated
or
what
are
the
particular
common
interfaces
the
node
needs
so
rather
than
users
being
exposed
to
all
the
intermittent
representations
directly
and
having
to
carefully
document
those
in
our
tutorials,
where
they've
been
quickly
fall
out
of
date,
because
you
know
the
Ross
graph
has
a
Ross
info
topic
that
that's
what
Ross
uses
to
inherently
announce
what
interfaces
are?
E
Maybe
you
have
to
take
in
account
all
these
weird
strange
permission
changes
between
our
mws.
You
should
really
push
the
effort
like
a
graphical
interface.
That,
then,
is
that's
what
we
document
and
then
all
these
abstractions
are
so
the
hidden
away
and
we're
less
likely
to
encounter
end
user
issues
where
they
get
confused
lost
in
terms
of
the
a
disconnect
between
implementation
and
documentation.
D
But
yeah
I
would
like
to
touch
upon
that
as
well.
The
and
what
you
were
suggesting
before
is
something
using
something
like
the
mde.
The
model
driven
engineering
tools
with
nodiel,
so
you
just
create
the
blocks
and
the
tool
automatically
generates.
You
know
the
other
rules,
because
otherwise,
as
Ruffin
was
saying,
it
may
be
challenging
to
have
something
correct.
D
If
you
remember
we,
we
stumbled
upon
the
same
problem
when
we
were
discussing
during
florential
presentation,
this
the
nodial
have
no
way
of
being
verified
and
that's
why
we
stress
about
the
graphical
user
interface
and
some
introspection
tools
for
the
graph.
Since,
in
some
of
our
previous
work
with
Ruffin,
we
were
applying
formal
methods
to
Discovery
data
from
DDs,
and
we
found
some
problems
in
the
way
policies
were
generated
because
we
were
verifying
directly
on
the
graph
rather
than
theological
aspect
of
the
system.
C
D
A
No
so
from
from
my
perspective,
there
is
no
one
solution.
That's
going
to
be
perfect
right.
You
can
listen
to
the
graph
and
miss
some
temporal
connections,
such
as
actions
or
all
the
likes.
You.
E
C
A
A
That's
not
perfectly
fitting
your
your
graph.
Now
we
all
we
are
discussing
about
the
same
thing,
alleviating
the
difficulty
of
writing
a
policy
for
a
given
graph,
and
we
are
both
I
mean
both
parties
proposing
a
different
solution,
but.
E
E
There's
there's
a
space
for
where
they
all
just
sort
of
integrated,
where
you
use
the
measurements
to
take
a
take,
a
raw
yeah,
a
run,
accurate
sample
of
what
are
the
interfaces
on
the
graph
topology
and
then
that,
like
that,
gets
back
ported
into
what
no
dead
node
node
DL.
Is
you
like
processing,
because
the
thing
that
no
deal
will
know
is
like
which
nodes
were
started,
and
you
know
how
the
nodes
are
composed,
and
so
it
can
take
the
the
particular
interfaces
that
nodes
have
been
using.
E
Do
the
association
on
which
interface
is
being
used
by
which
process
and
then
probably
then
it's
the
best
fit
and
then
maybe
no
deal
has
the
best
abstraction
on
on
how
to
communicate
the
particular
permissions
and
then
sros,
then
later
yeah,
it's
going
to
use
that
to
boil
it
down
back
again
to
the
mission,
so
we're
kind
of
doing
this
just
around
the
bush
kind
of
Avenue.
E
But
that's
the
way
it
probably
handled
the
abstraction
that,
based
on
the
communication
layer,
we're
using
it
definitely
seems
like
yeah
they
can
they.
They
work
together
in
that,
in
that
sense,.
C
So
I
I
just
wanted
to
add
my
cheer
students
in
here
and
I
agree
with
what
both
has
been
said
by
actually
the
previous
three
of
you,
Jen,
Luca,
Jeremy
and
and
Robin
I.
Just
want
to
add
my
peers
in
here.
C
Maybe
taking
a
couple
of
steps
back
and
and
sharing
with
the
group
some
of
the
observations
that
I've
that
I've
taken
and
that
some
of
my
other
colleagues
in
ads
robotics
take
pretty
much
every
day
when
we
speak
with
customers,
real
industrial
players
in
Industry,
considering
Ross
and
and
the
reality
is
that
designing,
complete
security
policies
remains
a
real
challenge
with
worlds
too
I'm,
not
just
speaking
about
what
using
one
or
two
of
the
tools
that
this
was
to
Features,
but
overall,
designing
these
security
architecture,
the
complete
one
or
multiple
security
policies
to
partition
properly
a
robotic
application
and
Frankly
Speaking.
C
While
writing
this
paper,
even
the
terminology
needs
to
be
polished
and
improved.
If
you
map
all
of
the
public
documentation
about
security
throughout
the
raw
space,
you
would
find
even
inconsistency,
something
we
we
ourselves
face
throughout
our
discussions
while
putting
together
this
this
content.
So
so,
overall.
C
To
be
honest,
I
I
was
really
happy
to
see
this
happening
discussion,
an
early
discussion
about
our
contribution,
because
that
fulfills
our
initial
objective,
which
is
awareness,
that's
what
we
wanted
to
somehow
push
forward,
awareness
and
consistency
or
what
concerned
a
security
talk.
C
So
I
am
I,
am
excited
about
that
and
I
just
want
to
hint
about
this
fact
that
there
needs
to
be
much
much
more,
that
we
need
to
do
I'm
personally,
very
excited
about
no
deal
and
the
opportunities
I
I
agree
about
the
fact
that
maybe
no
DL
can
set
up
a
base
template
template
that
then
can
be
improved
by
real
data
measured.
Well,
the
graphs
are
evolving
and
running
and
Frankly
Speaking.
C
C
While
reading
the
question
that
I
think
rodentia,
you
prepared,
which
was
I,
think
that
the
last
question
reads
us
something
like
like
how
can
nodiel
contribute
in
a
way
to
this
ongoing
proposal
and
framework
and
methodology
overall
well
I
mean
we
are
totally
open
to
actually
enhance
the
current
article
and
add,
maybe,
as
part
of
the
use
case,
that
was
studied
an
odl
extension
actually
I
would
I
would
love
to
review
such
a
pull
requests
contributing
to
this
repo
we've
been
contributing
for
a
few
years
now,
since
2019,
with
the
use
case
on
the
third
level,
three
and
and
Frankly
Speaking,
there's
even
academic
opportunities
there,
because
to
be
very
honest,
the
paper
was
submitted
to
iros,
but
reviews
are
still
to
come
and
maybe,
as
part
of
the
reviews,
you
guys
know
that
sometimes
three
submissions
or
or
even
the
reviews
for
final
submission
include
some
additional
contributions.
C
And
Florencia
you
guys
can
squeeze
a
bit
of
your
time
and
and
make
a
contribution
to
this
repo
that
we've
been
using
and
I
can
I
can
fetch
the
link
for
you
now
if
you
need
it,
but
it
has
been
public
for
a
few
years
now.
C
So,
if
you
can
add
some
bits
over
there
about
No
Deal
on
how
it
can
add
actual
value
which
I
think
it
does
to
the
current
proposal,
I
think
that's
that's
the
sweet
spot
we
want
to
reach
because
Frankly
Speaking
at
the
end
of
the
day,
the
major
complaint
that
I
keep
hearing
from
users
is
that
there's
simply
just
not
enough
security
talk
and
that
that
people
is
kind
of
like
waiting
to
start
believing
that
Ross
to
is
secure,
and
this
is
what
we
are
trying
to
convey
over
and
over
that
there
are
definitely
ways
to
ensure
security
at
the
roster
level.
C
A
We
at
canonical,
we
are
finishing
currently
our
2204
cycle
and
in
in
about
a
month
we'll
start
a
new
cycle
and
in
this
cycle,
I
do
hope.
I
do
expect
to
be
able
to
allocate
a
good
chunk
of
time
for
the
implementation
of
the
reference
robot
right,
I'm,
keeping
this
generic
and
open.
This
can
mean
a
lot
of
things
in
in
reality
in
terms
of
workload,
but
the
end
goal
at
least
would
be
would
be
very
clear.
We
want
to
have
a
reference
implementation
of
Ross
to
a
secure
robot.
A
B
C
So
I
know
I
know
just
quickly
responded
to
that.
I
know
you
guys
are
looking
into
the
third
level
four
and
and
I'm
entirely.
Okay
with
that,
I
just
want
to
hear
maybe
drop
a
comment
with
regard
the
fact
that
so
this
this
report,
that
I
just
shared
the
terrible
three
demo
which,
if
I'm
not
wrong,
was
originally
started
by
Raffin
and
Mikhail
and
then
later
on,
contributed
by
various
of
us
over
the
years.
C
There's
a
significant
amount
of
effort
in
many
hours
behind
this,
and
it
took
lots
of
lots
of
effort
to
reach
this
level,
and
it
is
still
taking
lots
of
effort
to
maintain
it
as
far
as
I
know,
so
so
I
guess
the
reason
why
I
was
suggesting.
That
is
because
this
sounds
like
the
lowest
hanging
fruit
to
to
Really,
bring
back
to
the
community
value
for
what
concerns
security
set
of
recommendations
well,
including
no
DL,
whereas
I
guess.
C
My
fear
is
that
if
we
jump
into
a
new
platform,
it
may
take
I,
don't
know
sometime
for
us
to
land
into
the
same
situation,
but
I
I'm
personally
blown
up
by
how
much
startup
three
sorry
third
level
four
is
getting
in
terms
of
media
and
excitement.
So
you
guys
might
be
right.
Selecting
that
Target.
A
B
E
Being
internal
bot,
3
is
purely
I
think
just
because
the
simulation
assets
were
there
and
matured,
but
now
that
it's
sort
of
been
lapsing
in
terms
of
according
to
the
latest
version
of
Ross
I
I,
it's
it's
fine,
because
the
stacks.
What
we
really
want
to
demonstrate
is
the
this
land,
the
navigation.
E
A
And
that's
where
I
was
going
to
to
conclude
implementing
the
nodiel
for
the
turtle,
but
free
most
likely
would
would
be
shared,
would
be
sharing
75
or
80
percent
of
the
effort
we
still
have
that
for
because
they
are
both
healing
plain
walls,
packages
right
so
implementing
the
DL
for
I,
don't
know
deep
Drive
controller
would
potentially
benefit
to
both
or
implementing
the
DL
to
the
navigation.
Navigation
stack
would
benefit
to
both
and
The
Wider
community.
E
So
so,
stepping
back
or
vectors,
like
maybe
the
big
bigger
picture,
did
anyone
have
any
comments
or
commentary
on
the
on
some
of
the
other
more
problematic
paradigms
that
are
inherent
in
Ross,
like
one
of
the
things
is
if,
if
a
robotics
designer
wanted
to
take
security
really
seriously
in
terms
of
like
information
flow
control,
you
know,
let's
say
that
that's
a
robot
manufacturer
OEM,
they
have
certain
liability
constraints,
but.
B
E
Integrating
with
the
consumer-based
subsystem,
obviously
there's
going
to
be
some
overlap
in
terms
of
where
the
consumers
you
know
product
sits.
On
top
of
the
of
the
robots,
OEM
and
they're
going
to
have
to
share
you
know
certain
topics
or
information
space
like
you
know,
transformed
trees.
Being
you
know
the
Pinnacle
example,
you
know,
one
of
the
things
is
that
transforms
are
inherently
you
know
necessary
in.
A
E
In
Nebraska
ecosystem
and
they're
usually
conveyed
over
like
single
topic
or
maybe
two
topics.
You
know
T
slash,
TF
or
slash
DF
static,
and
you
could
you
know
provision.
You
know
only
read
permissions
for
maybe
the
client,
but
you
know,
let's
say
the
client
sticks
a
robot
arm.
On
top
of
that,
then
they
need
right
permissions,
and
so
then
you
have
this
really
weird
overlap
of
where's
the
source
of
Truth
and
nothing's
really
restricted.
E
Based
on
the
the
Integrity
authenticity
of
the
transforms
being
published
over
these
monolithic
transform
topic,
you
know
I
think
that
maybe
there
may
be
a
case
there
that
well
maybe
they
should
just
be
on
different
DDS
domains
and
you
should
be
using
Bridges
and
stuff
like.
E
Know
is
there
a
more
elegant
way
that
raw
should
be
approaching?
You
know
security
at
that
granular
level,
or
is
that
just
is
that
inherently
out
of
scope?
Maybe
that's
something
some
of
the
questions
I
had
in
terms
of
what
we
should
be
doing
forward
in
terms
of
Security
Group
in
the
Ross
community.
D
B
E
A
E
That's
not
the
case
I've
seen,
there's
there's
like
a
autonomous
cleaning
robots
right.
So
this
is
this
is
something
there's
a
startup
out
at
San
Diego
and
they
kind
of
retrofitted
a
floor
cleaning.
You
know
stand
on
platform,
that's
kind
of
manned
and
made
it
unmanned
and
at
the
same
time
they
also
use
they
allow
third-party
Integrations.
Where
you
know
you
can
stack
on
a
surveillance
system.
So
now
the
the
floor.
Cleaning
robot,
is
also
a
security.
E
You
know
monitoring
system
Security
in
the
sense
of
just
like
Personnel
and
surveillance,
so
that
that's
a
sense
where
you
know
you're
Leasing,
the
OEM
is
sort
of
leasing
the
platform
and
the
platform
could
hurt
someone.
You
know
if
it
got
ran
over
a
child
or
whatnot,
but
it's
not
completely
decoupled
and
that's
also.
E
With
you
know,
various
add-ons
that
the
end
consumer
is
adding
on
top
of
the
platform,
so
if
the
entire
platform
ends
up
running
over
a
child,
is
it
the
third
malicious
third-party
you
know
add-on?
That
was
that
was
hacked?
Are
they
at
fault
or
is
it
the?
Is
it
the
OEM
that
made?
E
You
know
their
drivetrain
unsecure
because
someone
is
able
to
change
the
transform
tree
that
the
map
you
know
change,
so
it
collided
with
a
Personnel,
so
I
I
know
that's
kind
of
Niche
that
that
particular
example
but
I
I,
think
extrapolates,
where
you
know
you
have
like
clear
path
as
designated
specifically
like
selling
like
these
base
platforms
that
then
people
build
on
top
of
as
an
ecosystem,
and
it's
kind.
B
A
Right
and
that's
where
I
was
going
at,
if,
if
you
buy
your
platform
with
some
software,
you
you
can
either
have
a
tight
integration.
Your
your
extra
application
runs
on
the
same
Ros
graph,
in
which
case
you
do
have
access
to
well
to
the
entire
integration,
and
you
should
be
adapting
the
the
policies
and
the
whole
security
story
to
take
into
account
that
you
have
you
yourself,
the
Handover,
the
entire
graph.
A
A
If
you
are
adding
a
simple
security
camera
to
an
existing
robot,
that's
already
fully
autonomous,
then
maybe
you
don't
have
to
to
integrate
to
the
pre-existing
gross
graph.
You
know
what
I
mean
you
may
have
two
different
graph
and
if
the
first
graph
provided
by
the
Audi
odm
is
already
secure,
then
you
only
really
have
to
secure
yours.
E
So
so
the
the
you're
saying
that
we
should
really
just
the
ecosystems
really
just
rely
on
gateways
to
declassify
or
endorse
information
across
these
security
barriers.
A
E
C
And
nevertheless,
I
think
the
the
Baseline
discussion
in
here
points
out
towards
the
fact
that
having
more
use
cases
is
going
to
be
a
beneficial
aspect
for
the
working
group
to
to
be
reasoning
about
more
and
more
complex
Integrations,
so
on
that
I
second
I.
Definitely
second
Jeremiah,
with
your
efforts
on
on
trying
to
bring
this
to
to
more
and
more
robots,
so
I
think
that's
that's
going
to
to
add
value
to
it
and
hopefully
they'll
trigger
more
complicated
and
complex
discussions.
C
So
yeah
looking
forward
to
that
so
so
maybe
and
that
I
think
brings
it
to
the
second
question
that
I
think
Corinthia
you
prepared
about.
How
can
we
actually
maybe
apply
part
of
our
findings
or
hypothesis
to
some
of
the
ongoing
work
on
on
securing
robots
right?
Did
I
get
it
right
or
maybe
I
just
misinterpreted,
yeah.
B
Correct
so
I
was
actually
going
to
mention
in
the
interest
of
time
we
have
like
10ish
minutes
if
you
wanted
to
chant
to
the
next
discussion
so
and
we
touch
on
the
third
one
already
on
no
DL.
So
this
will
be
sort
of
our
last
question
to
touch
him
and
yeah.
That's
exactly
it.
B
As
Jeremy
mentioned,
we
are
prioritizing
development
of
the
reference
robot
in
the
next
few
months
in
our
roadmap
and
that's
a
joint
work
with
the
group
so
I
personally,
when
I
sorry,
when
I
saw
when
I
read
the
paper,
I
saw
a
lot
of
value
for
integrating
with
a
reference
robot,
especially
because
well
you're,
also
adding
this
this
framework
as
a
guidance
to
its
depths
that
cops
Inspire
framework
to
secure
a
robot,
besides,
obviously
promoting
the
technology
itself.
B
So
I
was
curious
to
to
heal
how
you
think,
sorry,
how
you
think
it
might
be
also
incorporated
into
a
plan
for
reference
robot.
Maybe.
C
That
is
I
think
that
is
a
great
question
and
indeed
opportunity
so
I
think
there's
started
translation
of
what
we
propose
to
the
actual
security
process
of
a
new
robot,
and
in
this
case,
while
tackling
the
title
load.
Four,
because
you
guys
are
attacking
the
title
of
four,
not
the
I
create
base
platform,
correct.
A
C
Base
yeah
yeah
I
mean
it's
totally
the
base,
but
I'm,
guessing
that
clear,
path's
going
to
add
a
few
more
things,
otherwise
yeah
so
yeah.
So
so
with
that
in
mind,
floating
I
would
say
that
there's
like
most
of
the
actual
approach,
like
section
three
in
the
paper,
is
directly
applicable,
especially
in
the
spin
and
I
would
say
in
a
chronological
manner,
in
the
sense
that
you
start
actually
doing
proper
modeling.
C
Then
you
connect
that
to
actually
determining
how
your
policy
should
look
like
by
like
establishing
the
right,
authentication,
authorization,
permissions
and
and
and
somehow
transferred
that
down
in
the
right
manner
to
the
right
mechanisms
that
is
worth
to
allow.
Then
it
comes
down
to
the
Generation,
and
that
is
something
that
definitely
deserves
more
attention
and
contributions,
because
right
now,
generation
of
distribution
of
the
policies
is
mostly
manual,
and
that
is
okay.
C
If
you
have
just
one
I
guess:
compute
system
within
your
robotic
system,
but
the
moment
you
have
like
multiple
socs
or
distributed
over
a
network,
then
it
becomes
a
bit
more
complicated.
So
so
that
is
something
definitely
that
maybe
might
lead
to
contributions
and
all
the
way
down
to
deploying
things
not
just
generating
them,
deploying
them
and
then
doing
constant
monitoring
to
feed
it
back
then
to
any
further
improvements,
while
modeling
things
again
over.
C
So
one
of
the
ideas
we
had
discussed
within
this
group
of
authors,
we
we
we
aimed
to
submit
this-
was
how
could
we
actually
transfer
this
into
the
existing
ross2
documentation?
And
maybe
that
is
something
you
guys
can
comment
on
as
the
working
group
leads.
So
what
is
the
best
way
to
bring
this
down
to
the
community
is.
Is
that
by
bringing
it
as
a
new
project
to
your
working
group,
called
I,
don't
know
methodology
for
securing
rods
to
computational
graphs?
A
A
A
That's
the
current
default
when
there
is
a
better
solution,
you
can
revisit
that.
But
at
the
moment
that's
that's
the
central
point
of
the
documentation
to
answer
your
your
question
as
to
how
what's
the
best
way
to
bring
that
to
to
the
community,
I
tend
to
think
that
it's
simply
a
tutorial.
You
know
a
step-by-step
tutorial
where
you
try
to
secure
a
small
but
compelling
example
of
a
Ross
stack
and
you.
B
A
D
On
this
on
this
end,
actually,
just
to
give
you
an
idea,
the
talker
listener
example
on
sros,
it's
broken.
It's.
E
D
If
you,
if
you
take
a
look
at
the
documentation
and
if
you
try
to
run
the
sros
command
directly
on
the
packages
of
the
demo
nodes,
the
artifacts
and
the
policies
that
are
generated
are
wrong.
They're,
not
correct
nice.
D
The
documentation
is
on
rolling
as
not
on
Galactic
the
under
on
their
Wiki
and
yeah.
I
have
to
open
the
the
issue
and
I
was
speechless.
I
have
to
say
that.
C
We
want
to
convince
industry
regardless
of
the
marketing
we
do,
regardless
of
how
great
tutorials
we
write.
But
if
we
want
to
convince
industry
that
ros2
is
secure
and
can
be
made
secure,
then
we
need
to
really
start
adopting
this
holistic
approach
when
we
start,
maybe
as
a
group
and
as
individuals
put
maybe
more
emphasis
on
on
things
like
what
gen
Luca
is
highlighting
things
such
as
what
Jeremiah
Jeremy's
is
headlighting
as
well
and
and
many
other
pending
aspects
about
communication.
I.
Think
writing.
C
Tutorials
is
definitely
something
fantastic
and
we
are
counting
on
you
guys
from
canonical
or
maybe
pushing
the
edge
on
that
because
you
have
tons
of
marketing
experience
there,
but
to
be
honest,
I'm
increasingly
receiving
this
input
about
the
fact
that
it
is
more
and
more
challenging
to
convince
people
that
Ross
can
be
made
secure
and
it's
not
it's
not
because
open
robotics
doesn't
advertise,
it
is
secure.
C
Certainly
they
are,
but
because
there
are
two,
the
actual
security
attitude
doesn't
really
match
with
how
how
other
Industries
are
approaching
Security
in
the
sense
of
policies
of
conventions
and
and
so
on
and
so
forth.
So
so
I
guess
and
and
maybe
I
diverged
a
bit.
But
this
is
very
connected
to
my
daily
work
and
and
to
my
daily
interests
and
I
think
it
it
somehow
can
be
tackled
in
a
way
by
addressing
properly.
C
This
second
question
Florencia
that
you
posting
here,
which
is
by
applying
a
systematic
framework
or
methodology
and
showing
the
way
and
showing
maturity
from
from
let's
say,
methodology,
perspective
manner,
because
if
we
just
recommend
people
that
adding
security
with
Ross,
it's
just
about
using
sros2,
then
nobody's
going
to
take
it
seriously
and
and
I'm
again.
I'm
confronting
this
in
serious
security
conferences,
where
literally
people's
asking
really
good
questions.
C
So
so
I
guess
maybe
embracing
more
this
methodological
approach,
while
of
course,
Very
connected
to
the
devsecop's
overall
movement,
which
seems
to
be
the
common
Trend
right
now
in
security
is
definitely
a
way
forward
and
Floridian
I'm,
hoping
that
our
writing
in
here
shed
some
light
into
how
to
start
applying
this
methodology.
If
it
doesn't
feel
free
to
reach
us
out
or
ask,
and
we
may
have
some
spare
Cycles
to
actually
help
you
guys
set
it
up
in
the
title
world
four.
C
For
now
it
is
set
up
in
the
third
level
three
and
it
is
working
as
far
as
we
have
tested
so
hopefully
that
that
also
paves
the
way
for
for
future
extensions.
C
C
Yeah,
nobody
mentioned
it.
Sorry
it
was
written
in
the
in
the
minutes.
At
the
end
of
the
second
bullet
point
big
bullet
point:
it
was
written
or
contribute
to
other
working
group
projects.
B
This
is
this
way
to
to
write
it.
It's
referring
to
the
secure
reference,
robot
or
other
projects
of
our
own
of
our
own
working
group.
Sorry,
the
missing
s.
B
Yeah
yeah,
like
you
mentioned
this,
this
framework
is
very
intuitive.
It's
very
straightforward
to
apply
to
any
project
really
and
the
way
I
saw
it
is,
is
a
way
to
introduce
developers
who
are
still
not
using
security
at
all,
but
you
may
find
accessible
to
think
in
adeptical
way,
which
is
you
know
it's
it's
graphical.
It's
like
very
a
set
of
very
clear
steps
that
make
sense
in
a
new
security
setting,
so
I
don't
think
it's
going
to
be
difficult
to
incorporate
it.
I
guess
it's!
B
The
question
was
more
like
how
how
this
framework
is
going
to
be
pushed
or
promoted
for
for
adoption.
Just
like
you
mentioned,
I
mean
the
discussion
is,
is
present
in
in
conference
and
all
over,
and
that
might
be
a
good
way
to
get
into
or
to
get
more
developers
to
start
applying
such
Frameworks.
B
C
So
so
I
think
I
think
that
is
very
interesting.
What
you've
just
said
and
I
guess
it
takes
us
to
actionable
aspects
right
away
right,
which
I'll
bring
right
now.
The
first
one
is:
how
can
we
bring
this
work
that
we
have
published
or
aim
to
publish
to
an
actual
project
in
the
security
working
group
and
I
defer
to
you
Florencia
and
to
you
Jeremiah,
to
try
to
activate
this
I
I
can
commit
to
put
some
of
my
personal
time
if
we
can
bring
finally
one
more
project
to
the
security
working
group.
C
Besides
the
the
sros
tooling,
you
guys
know
or
or
you
can
check
the
passport
requests
that
I've
been
pushing
to
get
more
projects.
I
would
just
say
once
again
I
think
we
need
more
projects.
We
need
to
get
more
people
excited
and
also
encouraged
to
contribute
in
one
way
or
the
other.
It
cannot
be
just
one
be
one
single
Direction
and
then
I
would
say
Florencia
to
your
words,
which
I
fully
agree
with
the
fact
that
it
aligns
with
the
devsec
Ops
and
the
way
it's
phrased.
C
It's
not
just
about
it's
not
just
about
introducing
non-security,
aware
people
into
it.
It's
also
about
conveying
the
right
message
to
security,
aware
people,
the
message
we're
in
security
is
not
really
a
product.
You
know
that
you
license
or
acquire
and
then
you're
secure,
good
to
go
come
on.
Security
is
a
process.
C
It
is
really
a
process
wherein
collaboration
between
security,
researchers
and
many
other
stakeholders,
including
manufacturers
and
end
users,
is
a
must
it's
needed
down
the
road,
and
this
is
really
not
being
conveyed
in
the
community
and
somehow
you
know
he
then
or
avoided
in
many
discussions
that
we
see
publicly
even
in
roscom's.
C
Let's
be
honest:
when
was
the
last
real
security
talk
that
happened
in
in
in
roscon
last
year
there
were
very
little
security
talks
the
year
before
that
very
few
security
talks,
and
that
actually
gets
me
to
something
that
excites
me
very
much,
which
is
asking
for
a
security
talk
this
year,
please.
So
if
there
is
anyone
actually
excited
about
doing
a
security
talk
with
us,
because
I
know,
Jen,
Luca
and
rafin
are
excited.
Please
reach
us
out.
Reach
me
out.
C
If
you
want
Florencia
and
Jeremiah,
if
you
guys
are
up
for
adding,
maybe
that
was
what
I
was
hoping
if
you
guys
are
up
for,
maybe
you
said
75
of
an
overlap.
Well,
that
means
that
it
might
not
take
too
much
to
add
then
no
DL
to
the
existing
tb3.
C
So
if
you
guys
were
willing
to
maybe
take
that
step,
I
think
it
would
be
exciting
to
partner
all
of
us
together,
and
maybe
you
know,
bring
a
joint
talk
for
Roscoe's,
but
yeah
I
would
just
say:
there's
a
lot
to
do
and
I
would
love
to
see
more
projects.
A
Right
so
to
quickly
answer
your
your
points
about
roscom,
that's
precisely
what
we
are
targeting
with
the
turtle
box.
Four,
then,
whether
it
is
a
torque
or
demonstration
on
the
side,
a
series
of
lightnings
I,
don't
know
my
my
Omni
concern
is
what
what
will
be
the
content
of
this
presentation
and
what?
What
new?
What
novelty
can
we
bring
with
respect
to
previous
talks?
But
definitely
that's
that's
what
we
are
targeting
with
that's
our
deadline.
With
this
reference
implementation,
ID
yeah.
We
want
to
have
a
fully
functional
security
robot
to
present
at
Roscommon.
A
A
B
A
So
I
would
make
two
small
comments.
First,
I'm,
just
throwing
you
guys
a
small
Peak
with
with
love,
of
course,
you're
not
mentioning
the
as
far
as
I
can
see
you're,
not
mentioning
the
security
working
group
as
a
as
a
working
group
in
your
in
your
paper.
A
Opportunity
for
bringing
people
along
you
are,
though,
giving
all
the
links
to
to
the
GitHub
repositories
and
whatnot,
so
no
doubt
that
the
reader
will
eventually
find
out
about
it.
A
The
second
thing
is
a
more
open
question
and
it's
not
meant
to
be
addressed
right.
B
A
Seem
to
more
or
less
agree
on
the
the
overall
path,
the
turtle
but
pretty
secure
for
what's
gone
and
all
the.
C
I
think
that's
a
great
suggestion
and
I
think
that
and
I'm,
just
speaking
here,
just
on
my
behalf,
but
I'm
I
presume
that
maybe
Ruffin
and
Angelica
might
be
in
agreement,
but
it
would
be
great
to
cooperate
with
more
of
you
guys
and
I
I
think
this.
This
is
exciting,
especially
if
we
add
more
and
more
value
to
it.
I
take
your
point
about
not
mentioning
directly
the
security
working
group,
Jeremy
I,
think
you're
totally
right
and
that's
it's
surprising.
None
of
us.
C
C
What
I
can
say,
though,
in
our
defense,
is
that
in
every
single
communication
we
made
across
social
media
and
communities,
we
definitely
try
to
bring
people
to
the
security
working
group,
but
I
take
that
as
really
really
positive
criticism,
because
we
should
improve
that
indeed,
and
on
the
on
the
what's
next
I
like
a
lot
that
suggestion,
why
don't
we
take
it
maybe
offline
or
to
the
chat
and
try
to
bring
some
ideas
on
what
else
we
could
aim
for
for
other
for
other
submissions,
but
in
terms
of
content,
I
would
just
recommend
everyone
to
have
a
look
at
the
conclusion.
C
A
section
there's
especially
the
last
paragraph
I,
think,
has
really
interesting
ideas,
so
those
of
you
managing
themes
or
having
resources
to
put
students
into
work.
Some
of
those
thoughts
we
believe,
are
interesting,
and
some
of
them
are
actually
really
low
hanging
fruits,
which
can
bring
significant
value
so
yeah,
yeah
well
said
I
would
say
Jeremy
thank.
B
A
Let's
take
this
talk,
this
discussion
offline
and
and
that's
it
well,
we
are
reaching
the
the
end
of
our
thank
you
all
for
for
attending
and
we'll
see
you
next
time
and
much
much
earlier.