►
From YouTube: IETF99-SAAG-20170720-1330
Description
SAAG meeting session at IETF99
2017/07/20 1330
https://datatracker.ietf.org/meeting/99/proceedings/
C
F
G
D
So
yes,
we're
starting
with
working
group
reports.
She's
very
we
know
well,
I
forgot
you
all
to
be
familiar
with
this.
If
you're
not
familiar
this
well
I
guess
you
can
read
it
or
read
it
later
right.
So
we'll
start
with
some
working
Barry.
Did
you
have
a
yeah
okay,
good
enough?
If
you
bring
them
in
order,
always
I
know
we
already
in
the
order,
turns
out
ace.
H
We
met
Monday
morning
and
we
have
lots
of
different
proposals
floating
around
and
how
to
carry
these
different
tokens
around
into
the
authentication
authorization.
So
we
had
lots
of
researchers
coming
along
and
providing
input.
We
unfortunately
didn't
make
that
much
progress
on
the
already
charted
items.
So
we
need
to
improve
on
that,
but
there's
a
lot
of
ideas.
C
I
So
we
met
Thursday,
we
actually
have
two
sessions.
Why
is
the
procession
and
then
the
formal
session?
So
this
time
we
had
lots
of
contributions
on
the
data
models
and
information
models,
so
this
meeting
mainly
is
to
make
sure
they
are
lying
with
each
other,
so
we
reached
a
really
good
result
on
that.
Thank
you.
Thank
you.
C
H
Yeah,
actually,
the
final
session
is
tomorrow.
So
in
the
first
session
we
talked
about
wrapping
up
some
existing
working
group
documents
and
we
also
had
Mike
Jones
present
a
best
current
practice
on
a
JWT.
After
we've,
some
researchers
found
attacks,
so
we
documented
those-
and
we
did
a
hum-
and
the
group
wanted
to
work
on
that
document.
So
that's
something
we
will
confirm
on
a
mini
list.
So
if
you
use
JW
T's,
that
may
be
a
good
document
to
look
at
or
if
you
implement
them.
E
Pgp
has
had
quite
remarkable
success
in
not
getting
anything
done
and
that
has
resisted
all
attempts
from
both
chairs
to
talk,
talk
to
people
offline
and
try
to
poke
things,
and
so
we
have
asked
Eric
to
close
the
working
group
whenever
he
is
ready
to
do
that
and
told
the
working
group
that
and
hope
that
that
might
spur
some
things
and
it
it
spurred
a
few
words
but
no
actual
action.
Eric.
Any
comments,
I
think.
E
D
K
The
as
the
co-chair,
this
is
dkg
I'm,
also
bummed,
that
we
didn't
manage
to
get
that
done.
There
is
continuing
work
in
that
space,
probably
and
if
I
can
manage
to
get
people
to
document
it
I'll
try
to
get
them
to
submit
drafts
and
if
they
actually
start
submitting
drafts
that
are
actually
functional,
then
maybe
I
will
come
back
and
ask
for
more,
but
we
have.
We
have
not
had
any
luck
in
getting
in
to
move.
We
don't
have
enough
implementers
in
the
group
right,
I.
Think.
D
D
C
C
G
C
C
L
Okay,
thank
you
so
in
the
dime
working
group,
so
we
have
the
base
protocol
and
initially,
in
the
previous
version
of
the
protocol,
we
have
an
end-to-end
security
staff
and
when
we
we
publish
a
new
version
of
this
one,
this
end-to-end
security
mechanism
was
deprecated.
It
was
based
on
CMS
and-
and
nothing
was
done,
so
we
are
still
looking
for
a
solution
for
end-to-end
security.
L
Basically,
it
could
be
between
edge
edge
agent
between
a
to
two
different
two
different
networks.
Anyway,
you
will
go
through
what
we
call
a
roaming
networks
or
you
have
a
bunch
of
agent
in
the
middle
so
or
you
may
have
end-to-end
real
end-to-end
communication
between
edge
two
clients,
client
and
server.
So
the
idea
to
have
a
solution
to
be
able
to
protect
a
VP's
next
slide
yeah.
L
So
the
status
is
that
the
end-to-end
security
requirements
is
now
an
RFC,
so
we
have
existing
requirements
coming
from
the
mobile
community
so
for
roaming,
especially,
but
also
by
operators
that
would
like
to
protect
their
diameter
signaling.
So
what
we
have
done
is
to
reuse
a
previous
draft
called
end-to-end
security,
but
no
work
was
done
on
this
one.
So
there
is
a
lake
lack
of
resources
and
also
like
of
expertise,
because
we
also
died
metal
part,
but
we
have
also
the
security
aspects
next
slide,
please
so
we
have
in
the
existing
draft.
L
We
have
a
basic
solution
where
it's
possible
to
have
integrity
and
confidentiality.
So
what
we're
defining
its
specific
AVP?
Where
were
you
able
to
sign
and
another
one
where
you
are
able
to
group
all
the
sensitive
EVPs
and
to
encrypt
them?
And
to
put
this
in
this
in
this?
What
we
call
a
group
AVP?
So
it's
it's
how
to
be
able
to
convey
to
convey
this
information,
so
the
solution
was
based
on
Jason,
so
using
the
Jason
Webb
signature
for
the
integrity,
protection
and
Webber
Kishin
for
the
confidentiality
we
were
thinking
about.
L
L
So
I
think
that
the
proposal
is
quite
simple
just
to
be
able
to
have
a
solution,
so
I
know
that
people
in
the
room
are
aware
of
this
issue,
so
we
I'm
looking
for
some
guys
that
could
help
on
this
work.
So
we
could
create
a
kind
of
a
design
team
and
so
on,
just
to
be
able
to
have
a
solution
done
so
any
any
volunteers
would
be
welcome.
So
you
can
contact
a
katrien
myself
and.
M
N
Phil
hon
Baker
I
had
to
generate
something
very
similar
for
my
meshwork
I.
It's
written
up,
but
not
separately
from
the
other
stuff,
but
I
can
pull
it
out.
I
did
notice
the
problem
with
you
with
applying
Jose
encryption
and
rolled
my
own
based
on
Jose,
but
doing
things
in
a
more
sane
fashion
for
doing
end-to-end
encryption
within
a
web
service.
So
so.
C
E
You
mark
met
this
morning
and
I
guess
the
main
thing
that
it's
discussing
right
now
is
whether
to
put
out
the
document
as
experimental
first
or
go
straight
to
standards
track
we'd
originally
been
figuring
standards
track,
but
Dave
Crocker
pointed
out
that
we
don't
know
the
operational
effects
of
of
what
we're
doing
and
could
use
an
experiment
with
that.
So
the
working
group
still
hasn't
decided
that
and
we'll
figure
it
out.
Otherwise,
it's
almost
ready
for
working
group
last
call
that's
the
arc.
E
O
Q
R
Lines
are
good
I
caught
you
a
quick.
We
are
monitoring
the
situation
in
TLS
with
interest.
C
S
Perk
is
getting
around
some
of
the
issues
it
had
with
its
double
context:
encryption
for
RTP,
that's
getting
to
a
point
where
it
would
benefits
from
some
more
review.
Well,
probably
in
a
couple
of
weeks
once
the
new
draft
is
out
would
benefit
from
some
review
from
folks
who
are
cryptographically
minded
and
have
some
enthusiasm
for
real
time
protocols.
So,
like
I,
said,
I
expect
there
will
be
a
new
version
of
that
double
encryption
draft.
In
a
few
weeks,
then
I'll,
maybe
ping,
the
cyclists
for
some
more
juice.
S
C
G
G
Okay,
talking
to
somebody
else
so
right
now,
we've
got
our
the
the
core
documents
that
we
were
charted
to
do.
They
are
in
they've,
been
overseas
for
a
while
we're
working
on
a
few
specs
related
to
the
best
best
practice
of
using
TLS
in
in
email,
various
forms
and
shapes.
But
after
that
we
can't
really
see
any
other
work.
So
if
the
security
community
has
stuff
for
you
ta,
you
should
bring
it
to
our
attention
before
Singapore,
because
after
that,
I
think
we're
gonna
be
done.
Thank.
R
T
C
R
C
H
Yep
for
deep,
we
had
a
tutorial
on
Sunday,
which
I
think
was
well
attended,
describing
on
how
Trust
song
works
and
and
the
scope
of
the
deep
effort
and
Wednesday.
We
we
met
to
discuss
the
Charter
and
I
posted
it
to
the
deep
mailing
list,
and
we
will
also
continue
with
the
webinars,
so
that
would
be
in
webinar
from
the
Intel
guys
about
the
trusted
execution
environments.
H
H
We
actually
met
today
and
sort
of
a
site
meeting
to
talk
to
those
who
are
developing
operating
systems
for
IOT
devices
were
the
potential
customers
or
the
potential
guys
who
are
impacted
by
this,
and
there
are
two
new
documents
out
there
posted
on
the
mailing
list
describing
what
the
standards
effort
could
look
like,
and
so
please
join
that
list.
If
you
care
about
firmware,
updates
and
or
IOT
devices.
Thank.
C
U
Sorry
I
didn't
realize:
I
was
on
the
agenda.
This
is
Sam
Euler
w3c
is
web.
Authentication
working
group
is
making
interesting
progress,
bringing
token
based
device
based
authentication
up
into
the
browser,
but
I
think
the
thing
they'll
be
of
more
interest
to
the
room,
is
we're
reaching
out
and
trying
to
get
more
eyes
doing
security
and
privacy
reviews
on
our
specs?
C
C
V
So
30
minutes
on
post
quantum,
cryptography
and
I
promise
you
there
are
no
equations
in
any
of
my
slides.
So
if
you
are
thinking
of
making
a
run
for
it
because
of
the
possibility
of
equations,
you
can
stay
stay
in
your
seats.
I'm
gonna
start
by
looking
at
something.
That's
manifestly
no
post
once
I'm
going
to
start
by
talking
a
little
bit
about
sha-1
and
talk
about
the
lifetime
of
sha-1,
so
we
go
back
to
deepest
darkest
history.
The
1990s
sha-1
was
published.
It
was
actually
a
tweak
of
a
design
from
1993.
V
There
were
various
attacks
on
Shia
zero
in
the
1990s
which
validated
that
switch.
Then
char
2
was
published
by
NIST
in
2001.
Nobody
used
it.
We
had
the
first
collision
attack
for
sha-1
and
2005,
except
it
was
an
attack
that
didn't
provide
any
collision.
It
was
an
estimated
complexity
of
2
to
the
63
hash
operations.
There's
a
famous
breakthrough
paper
by
wine
get
out.
It's
maybe
worth
remembering
that
wine
was
denied
a
visa
to
come
and
present
our
result
in
United
States.
At
the
time
things.
V
Some
things
don't
change
very
much
then
between
then,
and
now
there
were
lots
of
claims
and
counterclaims
about
improvements
to
this
attack,
and
you
know,
reactivations
of
the
complexity
in
2006
NIST
took
the
chef's
the
the
step
of
deprecating
sha-1
from
2010
onwards
for
all
federal
agencies
for
all
new
applications
requiring
collision
resistance
or
a
limited
set
of
applications
and
use
cases
in
2013,
Microsoft
gar
to
doing
something
about
showering
deprecation
from
2016
onwards.
For
new
code
signing
certificates
2014
still
no
collisions.
This
is
now
quite
some
time.
V
Nearly
10
years
after
Wang's
paper,
Wang
at
Al's
paper,
I,
should
say.
The
best
estimate
at
this
point
in
time
is
to
do
the
61
hash
operations
to
computer
collision
due
to
mark
Stevens
and
his
collaborators
in
2015.
Something
really
interesting
happened.
We
saw
something
called
three
star
collisions
which
are
a
precursor
of
food,
collisions
against
Taiwan,
not
quite
there
yet
requiring
moderate
computation.
Ten
days
on
a
GPU
cluster
with
sixty-four
GPUs
2017
sha-1
is
still
widely
used.
There's
lots
of
activity
in
the
C
browser
forum.
V
Maybe
some
of
you
have
seen
some
of
that
lovely
controller
state
happening
over
there.
Resisting
the
removal
of
sha-1
trying
to
extend
this
lifetime
is
gone
and
then
finally,
we
got
the
first
collision
since
sha-1
in
ferry
this
year,
I
joined
research
activity
between
the
researchers
at
CWI,
led
by
Mark
Stevens
and
by
Google
they
finally
actually
exhibited
sha-1,
collisions
q,
much
press
coverage
and
then
lots
of
researchers.
Like
me
saying.
Well,
you
knew
this
was
coming.
This
should
not
be
a
surprise.
V
V
Indeed,
here,
here's
a
related
graph
that
shows
the
uptake
of
sha-2
in
browser
trusted
certificates
on
the
on
the
public
Internet,
and
you
see
a
sha-1
search,
which
is
the
redline,
suddenly
started
decreasing
in
around
early
2014
and
then
sha-2
started
to
shoot
up.
An
md5
is
effectively
reached
you,
which
is
a
good
thing.
Anybody
want
to
guess
yeah
by
all
means
support,
but
let's
make
this
as
interactive
as
you
like,
except
please
don't
throw
things.
V
So
anybody
want
to
guess
what
happened
in
in
in
early
2014
to
make
this
thing
tick
up,
horribly,
the
first
vulnerability
with
the
logo,
not
the
last.
Only
the
first
okay,
so
I
think
what's
important
to
say
here
is
it
was
not
a
cryptographic
attack
that
led
to
the
switchover
from
sha-1
to
sha-2
in
certificates.
It
was
a
computer
security,
vulnerability
in
a
very
widely
deployed
piece
of
software,
open,
SSL.
V
Okay,
what
about
progress
in
quantum
computing
we've
seen
sometimes
something
happening
and
in
the
kind
of
classical
world
I
apologize
for
the
very
small
font
in
this
slide.
I
think
the
slides
are
online
if
you
want
to
have
a
look
at
them
and
follow
along.
So
there
were
so
many
things
to
put
in
here,
there's
been
so
much
progress
that
I
had
to
make
the
font
very
small.
So
it's
interesting
I
got
all
of
this
from
Wikipedia,
because
I
am
NOT
an
expert
on
on
quantum
computing
or
quantum
physics.
V
I
actually
missed
that
class
in
physics.
So
pre-1994
wikipedia
says
that
there
were
these
isolated
contributions
by
various
great
physicists
people
like
Stephen,
Fry's,
nah-ha
level,
Charlie,
Bennett,
etc.
Lots
of
interesting
little
nuggets,
but
nothing
no
kind
of
connecting
theory,
no
kind
of
concept
of
quantum
computing
as
a
thing
that
really
changed
in
94
and
then
96
because
of
Shor's,
algorithm
and
Grover's
algorithm.
So
short,
algorithm,
basically
breaks
discrete
log
and
factoring
problems
in
with
poly
many
gates
and
polynomial
depth,
polynomial
in
the
size
of
the
numbers
that
you're
working
with.
V
In
other
words,
all
public
key
cryptography
deployed
on
the
Internet
today
is
broken.
If
you
could
build
a
large-scale
quantum
computer,
they
could
implement.
Shor's
algorithm
in
Grover's
algorithm
in
1996
basically
gives
you
a
square
root,
speed-up
for
search
problems,
and
the
effect
of
that
is
to
have
the
effective
key
size
of
your
symmetric
primitives.
So
some
would
say
that's
why
we
have
aes
256
in
the
first
place
as
a
hedge
against
quantum
attacks
and
Grover's
algorithm
in
1998,
a
two
bit
quantum
computer
was
built
and
three
bits
using
NMR
in
2000.
V
We
got
to
five
bits
and
seven
bits
using
NMR
in
2001.
So
this
is
great.
The
number
15
was
factored
oh,
no,
no!
No!
No
seriously
on
Iran
a
quantum
computer
I'm,
not
I'm,
not
making
a
joke
here.
This
is
significant
progress
because
it
means
you
have
to
build
a
large
enough
computer
and
run
it
for
long
enough
to
find
out
that
15
is
three
times
five.
V
The
cube
eight
was
announced
in
2005.
If
I
have
no
idea
whether
I
know
it
means,
but
it's
good
good,
good
publicity
in
26
2006,
we
got
to
12
bits,
12
quantum
bits,
28
quantum
bits,
and
now
you
start
to
see
a
runaway
2008.
128
quantum
bits
this
stuff
just
got
real.
This
got
really
worrying,
except
this
is
d-wave,
so
there's
some
knowledgeable
laughter
from
the
front
for
those
of
you
further
back.
What
D
way
we're
building
is
a
quantum
annealing
machine,
so
it
implements
some
kind
of
annealing
algorithm
in
a
quantum
way.
V
It's
not
doing
general-purpose
quantum
computation
you
couldn't
use.
As
far
as
we
know,
one
of
these
T
wave
computers
to
do
to
run
Shor's
algorithm.
For
example,
we
do
know
that
Google
bought
a
dua,
wysz
een.
Anybody
from
Google
here
willing
to
confirm
that
and
then
tell
us
what
you're
doing
with
it.
I
have
been
told
that
they
dismantled
it
and
rebuilt
it
in
their
own
image,
so
some
how
to
do
their
own
stuff.
V
So
they
have
those
kind
of
people-
okay,
yeah,
true,
okay,
so
back
to
the
real
world
of
actual
general-purpose
quantum
computation,
fourteen
qubits
in
2011
and
then
in
2012,
the
number
21
was
factored
as
three
times:
seven,
then,
actually,
between
2013
and
2017,
there
wasn't
a
huge
amount
of
kind
of
publicly
announced
progress
in
factoring
and
I've
been
told
by
experts
in
the
field.
That's
because
factoring
is
the
wrong
problem
to
focus
on.
V
We
should
be
looking
at
other
things
than
factoring
I
think
it's
because
they
find
it
hard
to
scale
their
quantum
computers
beyond
the
number
of
bits
needed
here
there,
instead
focusing
on
more
fundamental
technologies
trying
to
get
them
in
place
before
attempting
to
fight
for
larger
numbers
in
late
2016
onwards,
the
physicists
did
a
pivot,
I
believe
is
called
the
pivot
in
the
tech
world
towards
focusing
on
something
called
quantum
suit
primacy,
which
is
actually
a
phrase.
I
find
personally
objectionable.
It
has
a
lot
of
really
ugly
overtones.
V
The
use
of
the
word
supremacy
here
and
what
this
means
is
building
a
large
enough
quantum
computer,
that
you
could
solve
a
problem
that
no
classical
computer
could
solve
in
a
reasonable
amount
of
time,
with
a
reasonable
amount
of
resources.
Think
of
gluing
all
of
Google
and
Akamai
and
CloudFlare
servers
all
together,
trying
to
solve
some
computational
problem,
not
being
able
to
do
it
but
being
able
to
do
on
a
quantum
computer,
something
like
some
kind
of
corns
from
chemistry
simulations
for
example,
and
so
now
there
are
rumors.
V
There
are
circulating
around
in
the
research
community
that
quantum
supremacy
will
be
announced
at
a
conference
in
Berlin
last
month,
Google
announced
that
they
would
achieve
quantum
supremacy
by
the
end
of
the
year,
but
I
think
that's
using
one
of
these.
Well,
who
knows
exactly
what
that
means?
Let's,
let's
not
speculate
too
much
in
2017
d-wave
launched
the
two
thousand
queue
which
has
two
thousand
quantum
bits
in
their
simulated
annealing
world.
That
reminds
me
of
the
story
of
Hewlett
Packard,
whose
very
first
product
was
called
the
Hewlett
Packard
200.
V
Okay,
because
it
sounded
good
to
start
with
something
that
had
a
large
number
and
a
small
number
it
looked
like
you
were
more
established,
I
guess,
D
Way
have
been
a
p-wave
have
been
around
for
a
while
now
IBM
unveiled
a
seventeen
bit
quantum
machine.
Sorry,
a
seventeen
qubit
machine
and
Google
and
Microsoft
Research
are
all
doing
cool
stuff
in
this
area.
So
there's
a
lot
of
energy
going
in
a
lot
of
money
going
into
this
progress
hard
to
judge
for
me.
V
So
what
I'd
say
is
that
to
draw
these
two
things
together
is
that
the
threat
of
large-scale
quantum
computing
is
kind
of
weakly
analogous
and
go
with
me
on
this.
It's
pretty
weak
analogy
to
the
threat
of
a
breakthrough
in
sha-1
collision.
Finding
and
the
breakthrough
might
be
imminent,
but
then
again
it
might
not,
and
we
know
do
actually
have
sha-1
collisions.
V
So
this
is
slight
really
should
have
been
updated,
but
for
a
long
long
time
people
were
speculating
well,
we
can
do
sha-1
collisions,
but
nobody's
done
it
yet
because
the
reward
is
not
high
enough
or
because
it's
actually
harder
than
we
think.
We
didn't
really
know
whether
it
was
going
to
be
possible.
No,
we
didn't
really
know
whether
those
algorithms
would
scale
or
not
to
the
size
required
it's
hard
to
quantify
the
risk
that
something
will
happen
and
hard
to
put
a
timeframe
on
it
happening,
but
meaningful
results
in
this
area.
V
We'd
have
a
substantial
impact
and
lots
of
smart
people
are
working
on
it
and
there's
been
a
tremendous
amount
of
research
investment.
On
the
other
hand,
maybe
quantum
computing
is
a
bit
like
fusion
research.
If
any
of
you
tracked
fusion
research
over
the
years,
it's
been
very
interesting,
there's
been
tokamaks
and
there's
been
no
kinds
of
plasmas
and
firing
lasers
at
pellets,
and
none
of
it
really
works
yet.
Okay,
so
here's
some
conversations
I've
been
party
to
with
people
in
the
quantum
computing
world.
V
Large-Scale
quantum
computing
is
only
a
decade
away,
but
he's
been
saying
that
for
25
years,
okay
in
terms
of
fundamental
physics
were
pretty
close
to
what
we
need.
There's
just
tons
of
engineering
work-
and
this
is
a
quote
from
Evan
Jeffery,
who
is
now
with
Google,
was
with
UC
Santa
Barbara
and
he
was
he
and
his
team
were
acquired
by
Google,
and
this
was
in
a
really
great
talk
that
he
gave
a
very
cold
head.
It's
sensible
talk
ly.
V
He
gave
about
a
quantum
computation,
a
real-world
crypto
in
2017,
and
you
can
find
that
on
the
real
world,
crypto
Google
as
for
a
YouTube
feet,
and
he
estimated
also
in
that
talk
that
to
break
a
1024
bit
RSA
modulus
would
require
something
like
250
million
quantum
bits
to
be
held
in
a
coherent
state
for
long
enough
to
do
the
to
do.
The
computation
I'll
just
remind
you.
The
current
state
of
they
are
is
17,
so
there's
a
gap,
but
apparently
it's
only
engineering
work.
V
Okay.
We
should
get
him
to
come
and
say
that
to
the
IETF
see
how
you
guys
react
to
that
yeah.
Okay,
so
is
there
a
coming
crypt
apocalypse?
I
should
never
put
a
word
on
a
slide
that
I
can't
actually
pronounce
I
find
apocalypse
very
hard
to
say.
I've
done
it
twice
correctly
now
and
I
think
I'll
quit
there.
We
don't
know
whether
there'll
be
a
scaling
breakthrough
or
not.
If
it
comes,
it
would
be
pretty
catastrophic
for
the
internet
and
for
internet
security.
V
So,
as
I
mentioned
before,
shor
Shor's
algorithm
would
mean
that
all
currently
deployed
of
a
key
cryptography
on
the
Internet
today
would
be
broken
without
exception.
Okay,
you
can
break
elliptic
curve,
cryptography
all
discrete
log
based
cryptography
based
on
finite
fields,
all
RSA
base
cryptography
and
that's
pretty
much
what
we
have
in
deployment
today,
interestingly,
because
of
the
key
sizes
and
the
sort
of
scaling
properties
of
quantum
computation
elliptic
curve,
cryptography
would
break
before
an
RSA.
V
So
lipton
of
cryptography
is
weaker
for
the
same
key
cipher
for
a
smaller
key
size
and
I,
see
and-
and
one
of
the
one
of
the
things
to
think
about
here-
is
that
an
attacker,
a
well
resourced
attacker,
think
nation-state
adversary.
Nsa
could
capture
trusting
okay,
you're
catching
up
good?
Okay,
a
nation-state
bursary.
Could
capture
interesting,
looking
difficult
exchanges
now,
okay,
so
from
my
email,
server
to
Eric's
mail
server,
where
we're
discussing
this
talk
and
what
am
I
going
to
say,
I
think
was
a
question
yeah.
Just
a
quick
question
on.
S
The
ECC
point
it
seemed
Richard
parts
is
this
been?
Is
this
been
a
point
with
that?
Is
it
gotten
actually
gotten
some
analysis?
Besides
the
hand
waving
keys
are
longer
need
more
cubits,
because
there
are
like
time-space
trade-offs
that
apply
here,
and
so,
even
though
you
might
be
able
to
get
by
with
fewer
cubits,
you
might
need
a
longer
coherence
time.
Is
it
similar
walls
I
think
a
detailed?
How
all
of
this
analysis.
V
W
Paul
happened
here.
There
was
just
a
paper
like
two
months
ago
on
this,
which
actually
says
that
maybe
ECC
is
going
to
be
almost
as
hard
for
equivalent
strikes.
The
last
thing
I
heard
was,
you
need
many
more
registers,
so
you
need
a
bigger
computer.
I
believe
the
answer
is:
there's
not
agreement,
yet
there
probably
will
be
agreement
on
this
within
the
next
few
years.
Thank.
V
You
very
much
for
that
clarification
poll,
okay,
I'm
as
I
said,
I'm,
not
an
expert.
So
if
people
here
know
more,
please
come
and
grab
the
mic
and
give
the
rest
of
the
presentation
that
would
be
really
great.
I
could
sit
down
okay,
so
we
really
expect
some
warning
of
impending
disaster.
We
would
hopefully
see
in
research
papers
published
research,
but
of
course
you
don't
necessarily
know
what's
going
on
inside
the
government
research
lab-
and
we
know
that
replacing
kryptonite
scale
takes
time.
V
Okay,
it
took
a
long
time
to
get
rid
of
max
and
encrypt
and
TLS
they're
still
hanging
on.
There
I
think
we
kind
of
killed
our
c4,
but
still
it
takes
time
and
you'll
still
see
it
in
either
normal
corporate
data
centers,
where
they
don't
really
keep
that
up
to
date
with
security.
Oh
sorry,
I
didn't
say
that
I
didn't
say
that
okay,
replacing
crypto
at
scale
takes
time
and,
of
course,
as
I
mentioned
traffic
capture
now
it
could
be
broken
later.
V
So
if
you've
got
data
that
needs
to
be
kept
secure
for
decades
and
you
think
there's
a
risk
of
large-scale
quantum
computation,
then
you
need
to
apply
some
kind
of
precautionary
principle
and
think
about
migrating
your
crypto
sooner
rather
than
later.
Okay,
there's
even
a
conference
for
this.
It's
called
katha
crypt,
it's
a
pretty
doom-laden
conference,
but
you
know
if
you
you
could
go
okay,
okay,
so
how
should
we
move
forward
from
here?
Well,
this
is
not
the
way
to
move
forward,
let's
think
about
post
quantum
cryptography.
V
Instead,
what
is
pqc
post
pontipee
of
geography
is
cryptography
based
on
conventional
techniques,
so
conventional
public
key
cryptosystems.
These
would
run
we
would
compile
on
your
computer
and
run
okay
on
a
classical
computer
in
a
reasonable
amount
of
time,
and
we
believe
to
the
best
of
our
knowledge
that
they
resist
the
known
quantum
algorithm,
so
Shor's,
algorithm
and
so
on.
So
this
area
is
called
pqc
and
sometimes
it's
called
quantum
safe
cryptography.
Sometimes
it's
called
quantum
immune
cryptography.
V
If
you
work
in
the
area
of
quantum
key
distribution
and
you're
a
physicist
and
you
want
to
embrace
post
quantum
cryptography,
then
you
call
it
quantum
safe
okay,
so
we
are
bringing
other
things
into
your
fold
and
extending
your
reach
and
so
on.
So
there's
a
little
bit
of
politics
here,
we're
going
to
call
it
post
quantum
cryptography
the
main
candidates
for
this-
and
this
is
what
I
would
have
had.
The
equations
had
I
been
braver.
V
Our
lattice
based
cryptography,
which
has
been
studied
for
twenty
years
now,
I
think
code
based
cryptography,
which
has
actually
been
studied
for
much
longer.
The
michaelis
code
base
scheme
uses
gompa
codes.
These
are
error,
correcting
codes,
and
it
was
already
in
the
late
1970s
early
1980s
being
proposed
and
explored
there's
newfangled
stuff
based
on
nonlinear
systems
of
equations.
They
have
a
pretty
patchy
history,
there's
been
a
lot
of
schemes
proposed
and
then
broken
we're
still
finding
out
what
makes
these
problems
hard
and
then
the
latest
greatest
thing
is
elliptic
curve.
V
Isogen
ease
if
I
was
our
tech,
investor
I
would
not
invest
and
let
the
current
isogen
ease
right
now
we're
extremely
new
and
we're
still
really
figuring
out
what
kind
of
security
they
offer.
So
there's
four
main
approaches
that
the
researchers
are
currently
exploring
and
these
different
approaches
they
could
be
vulnerable
to
further
advances
in
quantum
algorithms.
So
it
could
be
that
tomorrow,
Peter
shor
comes
up
with
a
new
algorithm
that
solves
hard
lattice
problems.
We
simply
don't
know.
V
What
we
do
know
is
that
smart
people
in
the
quantum
algorithms
community
have
been
working
for
a
long
time
to
develop
new
quantum
algorithms
and
so
far
with
one
exception.
We
have
no
applied
to
lattice
based
schemes
or
to
these
other
schemes,
and
the
exception
is
this
paper
by
Elder
and
shor
that
was
published
much
earlier
this
year
and
then
almost
immediately
withdrawn,
because
it
was
a
bug
in
the
paper.
V
It
was
a
very
unfortunate
thing,
because
this
was
the
first
time
that
somebody
of
Peter
short
stature
had
something
interesting
to
say
about
lattice
based
cryptography,
and
this
was
kind
of
a
wake-up
for
the
classical
cryptography
community.
This
was
a
very
interesting
piece
of
news.
It
turned
out
that
the
the
the
parameters
that
he
was
able
to
attack
using
his
quantum
algorithm
were
not
the
parameters
that
we
would
use
in
cryptography,
but
still
it
was
maybe
the
first
quantum
in
the
armor
for
lattice
based
crypto.
V
So
this
was
a
significant
thing,
but
as
I
say
that
they
were
was
withdrawn,
but
maybe
he'll
come
back
with
a
new
idea
with
working
with
Eldar
in
the
future.
We
just
don't
know.
There
was
also
a
paper
published
by
GCHQ
in
the
UK
now
renamed
as
NCSC
called
soliloquy,
and
this
was
a
explication
of
their
failed
attempt
to
build
a
lattice
paste
crypto
system
that
was
continum
in
they
found
a
quantum
attack
against
their
particular
scheme.
V
So
with
another
data
point
there
for
you
that
maybe
this
stuff
has
some
ways
to
go
before
we
really
understand
in
security
against
quantum
algorithms.
Even
the
conventional
security
of
these
schemes
is
not
well
understood.
In
all
cases,
we
have
a
pretty
good
handle
on
the
security
of
hash
based
signature
schemes,
for
example
X
MSS,
which
has
been
working
its
way
through
CF
RG
and
is
close
to
being
an
RFC.
V
Now,
there's
a
later
scheme
called
Sphinx,
which
is
developed
by
some
academic
researchers
a
couple
of
years
ago,
and
we
have
a
pretty
good
handle
on
how
to
choose
parameters
to
make
these
kinds
of
things
secure.
They
rely
on
pretty
standard
properties
of
hash
functions,
which
we
believe
are
not
really
susceptible
to
quantum
attacks.
We
even
have
some
lower
bounds
or
the
strength
of
quantum
attacks
against
these
things,
but
for
things
like
lattice
based
crypto
and
code
based
crypto
and
nonlinear
systems
of
equations,
the
academic
research
community
is
really
no
digging.
V
X
V
A
question
of
parameter
selection
and
yes
and
no
I
mean
there's
trade-offs
between
the
number
of
signatures
you
can
sign
and
the
size
of
signatures
and
the
cost
of
signing
and
so
on.
I'm
I
think
on
the
next
slide.
I'll
come
to
this
point,
talk
about
some
of
the
characteristics
of
these
schemes
and
maybe
maybe
I'll,
try
and
address
it
there.
Okay,
yeah!
Here
we
go
so
it's
true,
and
this
is
really
what
Steven
is
hinting
at,
that
the
current
post
quantum
schemes
that
we
have
are
not
as
performant
as
the
pre
quantum
schemes.
V
We've
all
got
very
used
to
it
to
curve
cryptography,
not
really
consuming
that
much
power
on
our
servers
or
you
know
not
that
much
bandwidth
on
our
networks.
So
typically,
what
you're
going
to
find
in
the
future
is
larger,
public
keys
and
larger
key
exchange
messages,
larger
cipher
texts,
etc.
I,
don't
have
hard
numbers
for
you
here.
We
can
discuss
that
later.
V
If
you
want
this,
this
field
is
still
very
much
evolving
and
on
the
flip
side,
you
actually
often
get
much
faster
cryptographic
operations
to
actually
do
an
encryption
with
some
of
these
lattice
based
schemes
is
just
a
matrix
multiplication
and
adding
some
noise
vector.
So
encryption
is
much
easier
than
some
kind
of
finite
field
exponentiation,
for
example.
This
is
good
news.
V
The
performance
may
suffer
as
we
refine
our
understanding
of
how
hard
the
hard
problems
are
in
this
area,
as
we
refine
our
understanding
of
how
to
choose
parameters
for
security,
so
better
attacks
which
will
emerge
over
time
as
people
tweak
their
code
and
come
up
with
new
ideas,
imply
that
larger
parameters
are
needed
for
the
schemes
and
that
can
also
lead
to
a
particular
approach
is
just
being
completely
abandoned.
So
we've
seen
that
a
lot
with
some
of
the
systems
of
equations
approaches.
V
People
had
ideas,
they
tweak
them,
you
tweak
them
interpret
them,
and
eventually
they
said.
Okay,
we
give
up.
We
can't
make
this
secure
using
this
current
set
of
ideas
and
parameter
selection.
This
area
is
a
much
more
complicated
question
than
it
is
for
RSA
and
ECC.
We've
got
used
to
magic
numbers
like
2
or
4
8
bits
for
RSA
or
3
or
two
bits.
If
we
were
crazy
about
security
or
two
five,
six
bits
for
ECC.
V
There
are
more
parameters
to
play
with
in
the
post
quantum
schemes
and
it's
harder
at
the
moment
to
figure
out
exactly
how
to
set
them
put
another
way,
we're
roughly
where
we
were
for
RSA
in
about
1982,
okay,
I
would
say:
I
wasn't
around
I
mean
I
was
born
then,
but
I
wasn't
doing
cryptography.
Then
I
was
just
out
of
short
trousers,
but
looking
back
in
the
literature,
there
was
a
lot
of
debate
about
exactly
how
big
or
ISA
should
be
our
RSA
number
should
be.
V
V
Okay,
I'm
gonna
speed
up
a
little
bit.
So
where
are
we
right
now
with
post
quantum
cryptography?
We
are
still
doing
research,
lots
and
lots
of
research
and
we're
progressing
from
research
toward
standardization
and
appointment
at
a
fair
rate
of
notes.
There's
a
couple
of
examples
here:
there's
something
called
the
Internet
defence
prize
which
is
awarded
at
using
execute
each
year
and
it
was
given
to
a
lattice
based
key
change.
V
V
It
was
a
limited
time
experiment
to
understand
the
impact
on
bandwidth
and
processing
time
and
there's
a
huge
amount,
a
really
an
increasing
amount
of
mainstream
crypto
research
going
on
at
the
moment
in
this
area,
but
we're
not
there
yet
and
on
the
standardization
site,
NIST
last
year
announced
a
process
for
standardizing
post
quantum
public
key
algorithms,
starting
announced
in
2016
the
closing
date
for
your
submissions.
If
you're
planning
to
make
one
is
November
the
30th,
2017
and
I
know
there
are
lots
of
academic
and
industrial
research
lab
teams
now
getting
their
proposals
together.
V
I
just
want
to
show
you
how
long
the
process
is
expected
to
run
for
so
thank
back
to
the
days
of
standardization
of
aes
and
add
an
extra
two
or
three
years
to
that,
and
then
at
the
end
of
that
we
may
have
a
portfolio
of
post
quantum
algorithms
that
we
can
use
or
nist
have
reserved
the
right
to
make
a
decision,
but
further
evaluation
is
needed.
Okay,
so
it's
going
to
take
some
time,
Haemon,
you're
thinking,
I've
heard
about
this
thing
called
quantum
cryptography
and
quantum
key
distribution.
V
Why
are
we
not
using
that
instead
of
post
quantum
cryptography?
Well,
let
me
let
me
destroy
quantum
key
distribution
that
destroys
too
strong
a
word.
Let
me
let
me
critique
quantum
key
distribution,
okay,
so
this
is
great
right.
This
is
the
physicists
will
tell
you.
This
promise
is
security
based
only
on
the
correctness
of
the
laws
of
quantum
physics
and
the
correctness
of
quantum
physics
is
not
in
doubt.
It's
been
verified
to
ten
decimal
places.
V
Maybe
ten
decimal
places
is
not
very
large
when
you're
thinking
about
security
parameters
for
crypto,
okay,
we
considered
an
attack
at
a
level
like
two
to
the
minus
120,
and
then
we
scrapped
your
block
cipher.
So
ten
decimal
places
doesn't
really
cut
it
anyway.
Okay-
and
this
is
often
contrasted
to
the
security
offered
by
public
key
cryptography.
We
always
said
that
public
key
cryptography
is
vulnerable
to
quantum
computers
and
and
quantum
key
distribution
would
not
be
etc,
etc.
I
got
some
quotes
together
here.
This
is
from
Brian
loans.
Dr.
V
Brian
loans
from
kinetic
quantum
cryptography
offers
the
only
protection
against
quantum
computing,
and
all
future
networks
will
undoubtedly
combine
both
through
the
air
and
fiber
optic
technologies.
Okay,
here's
another
example:
all
cryptographic
schemes
used
currently
on
the
internet
would
be
broken.
That's
not
actually
true.
This
was
said
by
you
bizarre
who
is
an
IEC,
our
fellow
and
ex
editor-in-chief
of
the
Journal
of
cryptology,
so
he's
a
pretty
solid
cryptographer,
and
he
said
that
too,
at
a
lunch
meeting
of
some
Institute.
This
is
an
early
example
of
alternative
facts.
V
Here's
MIT
technology
review
as
early
as
2003.
This
is
2td
is
one
of
10
emerging
technologies
that
will
change
the
world,
even
Burt
cholesky,
who
many
of
us
know
and
highly
respect
said
back
at
this
time.
If
there
are
things
that
you
want
to
keep
protected
for
another
10
to
30
years,
you
need
quantum
cryptography.
V
These
are
examples
that
I
gave
at
a
talk
10
years
ago
this
year
at
the
fuse
Institute
at
the
University
of
Toronto.
So
these
are
all
10
years
old.
So
what's
the
big
holdup,
why
we're
not
already
using
quantum
key
distribution
but
I'll,
give
you
four
reasons
and
I
have
backup
slides.
So
if
anybody
wants
to
debate
this
with
me,
let's
list
it
in
the
coffee
break
after
and
qkd
doesn't
actually
do
what
it
says
on
the
tin.
V
It
doesn't
distribute
keys,
it
expands
existing
keys
and
if
you
already
have
a
key
in
place,
you
can
expand
it
in
all
kinds
of
ways
and
build
as
many
keys
as
you
need
to
using
things
like
key
derivation
functions.
Ok,
qkd
has
significant
limits
on
both
range
and
rate
range
is
the
most
fundamental
one.
You
may
have
heard
about
the
Chinese
quantum
key
distribution
network
running
from
Beijing
to
Shanghai.
It
sounds
very,
very
impressive,
but
it
does
not
provide
end-to-end
security.
V
It's
a
sequence
of
links
join
together
and
each
intermediate
point
you
can
read
the
traffic
in
the
clear.
Ok,
if
you
want
end-to-end
security,
qkd
is
not
going
to
give
it
to
you
until
we
have
quantum
repeaters,
but
they
require
a
quantum
computers
and
I.
Think
I've
told
you
about
quantum
computers.
Already.
V
Security
in
theory
does
not
imply
security
and
practice.
In
theory,
they're
the
same
in
practice,
they're,
not
I,
think
Yogi
Berra
said
that
I'm
here.
The
idea
is
that
we
have
to
build
devices.
We
have
idealized
models
of
the
physics,
but
the
devices
will
have
weaknesses.
They'll
have
side
channels,
you
can
shine
laser
beams
into
single
photon
detectors
and
cause
them
to
go,
haywire,
etc,
etc,
etc.
V
Okay,
so
proving
secure
systems
is
different
from
designing
secure
systems
and,
finally,
quantum
key
distribution
does
know,
offer
significant
practical
security
advantages
over
what
we
can
currently
do
at
low
cost
with
conventional
techniques,
by
which
I
mean
symmetric
key
cryptography.
Ok
check
me
later.
If
you
want
the
details
on
exactly
why
that's
true?
Ok,
it's
not
just
me
this
is
this
is
my
government.
This
is
the
view
from
ncsc
or
GCHQ,
as
was
quantum
key
distribution.
V
They
actually
took
the
unusual
step
of
publishing
a
white
paper
about
two
years
ago
outlining
their
view
and
here's
what
they
say.
The
summary
is
that
Q
KD
has
fundamental
practical
limitations,
does
not
address
large
March
the
security
problem
and
is
poorly
understood
in
terms
of
potential
attacks.
By
contrast,
this
is
wonderful.
V
Is
his
cat
net
for
someone
like
me
by
contrast,
postpones
on
public
key
crypto
or
appears
to
offer
much
more
effective
mitigations
for
real
world
communication
systems
from
the
threat
of
future
quantum
computers,
in
other
words,
invest
in
P
Q
C,
naught
Q
key
T,
ok,
coming
close
to
the
end
now.
What
should
I
etf
do
so?
Cfr
g
is
on
this
to
some
extent
we're
looking
at
developing
internet
traps
for
hash
based
answers.
We've
got
two
drafts
in
development
at
the
moment.
V
One
is
very
close
to
being
finished
and,
as
I
said
before,
this
is
a
very
mature
and
West
risky
area,
but
I
think
that
all
the
other
post
quantum
schemes
are
still
in
that
kind
of
difficult
teenage
phase
right
where
they're
kind
of
okey,
you
know
kind
of
gawky
and
awkward
and
likely
to
lash
out
at
any
moment
so
and
that's
in
the
research
community,
nevermind,
standardization
and
actual
experience
of
deployment.
So
we
have
a
long
way
to
go
here
and
my
view
is
that
NIST's
process
is
where
the
action
is
going
to
be.
V
So
in
this
has
the
resources,
the
secretary
at
the
experience
needed
to
run
a
proper
process.
You
might
think
back
to
things
like
the
Julie,
C
fiasco,
and
maybe
NIST
is
not
the
body
that
we
should
trust
the
developer
crypto
standards
but
I
believe
that
isn't
a
significant
concern
here
because
of
the
very
heavy
involvement
of
the
academic
community.
We'll
have
I
hope
that
and
I
believe
that
the
academic
community
can
keep
the
process
honest
I'm.
The
scientific
experts
will
all
be
concentrating
their
efforts.
V
They
are
with
NIST
and
not
with
IETF
or
other
standards
bodies,
at
least
in
the
next
five
to
seven
years.
That's
where
the
action
is
going
to
be
so
I
think
my
personal
view.
This
is
known
as
CFR
G
view
or
anything
else.
Just
my
personal
view,
we
should
wait
for
an
s
processor
on
its
course.
We
should
be
ready
to
rule
over
once
those
algorithms
are
finalized.
V
V
We
should
also
maybe
do
some
work
and
see
if
our
G
could
do
this
understanding,
how
best
to
combine
pre
or
classical
crypto
systems
with
post
quantum
schemes
to
so
that
we
get
a
hedge
against
the
failure
of
either
type,
and
we
should
also
identify
and
resist
any
efforts
to
do
the
kind
of
sto
shopping
that
you
sometimes
see
where
people
go
to
multiple
standards
bodies
until
they
get
one
that
that
takes
their
stuff,
and
we
should.
We
should
allow
this
to
do
its
work
with
the
help
of
the
scientific
community.
V
Okay,
so
final
slide,
it
might
be
coming
or
it
might
not
be
I'm
not
going
to
try
and
say
that
word.
It's
possible
that
post
quantum
co2
is
a
massive
misdirection
of
the
scientific
community,
designed
to
distract
us
from
things
that
really
matter
like
we
could,
if
he
helm
and
parameters
or
I,
don't
know,
maybe
repeated
ephemeral
values
in
the
TLS
protocol
or
something
sigh
I
makes
me
so
sad
makes
me
so
sad
folks,
I'm
channeling
Donald
Trump
there,
and
we
can
so.
V
We
can
hope
and
I
believe
that
in
this
process
will
proceed
in
an
orderly
fashion.
They
were
actually
in
some
sense,
taking
their
cryptographic,
reputation
on
it.
They
have
to
get
it
right
and
they
have
to
put
adequate
resources
into
it,
because
they're
still
in
the
process
of
doing
a
repair
job
based
on
what
happened
with
Julie
C
and
that
whole
that
whole
fiasco
so
I
think
at
the
end
of
it,
we
will
see
a
very
sensible
and
conservative
portfolio
of
algorithms
of
options.
For
that
will
suit.
V
Itf
needs
I
mean
time
those
things
ietf
can
be
doing,
as
I
mentioned
on
my
previous
slide,
to
make
the
transition
as
smooth
as
possible
so
that
when
we
we
get
those
algorithms
for
an
S
we're
ready
to
start
using
them.
Okay,
so
that's
everything
I
wanted
to
say:
please
form
an
orderly
queue
of
the
mic.
I,
don't
have
to
be
happy
to
talk
to
you
thanks
a
lot.
S
See
thanks
video
you
can
is
Richard.
Barnes
I
will
+1
the
suggestion
on
your
previous
slide
that
we
that
CFR
did
you
take
a
look
at
this
question
of
how
you
combine
classical
and
post
quantum
algorithm
work.
Free
quantum
impose
commonalities.
If
you
will,
because
I
think
that's
gonna
be
critical
to
transition
and
I
think
it
was
one
where
we're
developing
a
solution
with
some
degree
of
generality
I
think
it
would
be
really
useful
in
facilitating
that
transition.
I
think
it
made
that
request
cfrt
before
I
make
it
again
now.
If
anyone.
S
Be
interested
I
was
wondering
if
you
had
any
thoughts
about
like
how,
if
there
are
ways
that
it
seems
like
you've,
kind
of
parsed
out
kind
of
in
this
process
and
I
ATS
I,
see
FRG
processes
and
fairly
separate
stacks
here.
There's
wonder
if
you
have
any
thoughts
on
how
they
could
maybe
support
each
other.
If
there's
things
that
see,
if
are
G
or
IETF
could
do
that
would
be
supportive
in
this
process
in
terms
of
the
different
perspectives,
evaluating
proposals
or
doing
Interop
testing,
or
something
like
that.
Okay,.
V
Yeah
thanks
for
that
suggestion,
that's
an
interesting
idea.
Maybe
we
could
have
some
ITF
folks
attending
the
NIST
workshops
that
will
take
place
and
feeding
in
requirements.
I
mean
the
sort
of
requirements
gathering
phases
somehow
passed
already,
there's
a
requirements
document,
that's
been
that's
been
published
already
last
year,
but
they're
probably
still
opportunities
to
influence
if
new
requirements
are
I,
so
I
encourage
everybody
here
to
go,
find
the
NIST
call
for
proposals
document
at
this
website.
Let
me
find
the
did.
I
go
back
past
it
already
I'm
trying
to
find
the
URL.
V
Well,
you
can
find
online
anyway
and
there
you
can
see
the
call
for
submissions
and
there
it
talks
about
these
kinds
of
requirements.
So
something
is
missing.
Then
ITF
people
should
should
get
involved
more
generally
and
I'm,
not
sure
off
the
top
of
my
head.
I,
don't
have
any
grace
for
how
we
can
we
can
compliment
what
they're
doing
I.
Aside
from
the
things
I
already
mentioned,
I
think
the
I
know
for
a
fact
that
all
of
the
key
players
in
the
research
community
in
this
are
really
focusing
their
efforts
on
the
noose
competition.
V
S
Want
to
come
back
yeah,
just
a
quick
I
think
that's
a
good
insight.
I
think,
maybe
that
we
have
a
little
bit
more
of
the
industry
players
here,
and
so
perhaps
this
could
be
a
way
to
make
that
connection
between
the
academic
researchers
and
doing
some
more
kind
of
operational
testing
and
in
my
systems
a
little
closer
direction.
Thank.
W
You
Paul
Hoffman
two
things
that
are
related
one
when
you
said
well:
I,
don't
know
how
big
the
keys
or
the
signatures
might
be
just
for
the
people
in
the
room.
Current
estimates
are
they're
going
to
be
at
least
an
order
of
magnitude
larger,
so
we're
talking
about
50,000,
bit
keys
and
suchlike
that,
depending
on
signature
sizes
and
such
obviously,
this
is
one
of
the
things
that
NIST
is
considering
on
trying
to
minimize
that.
W
But
if
you've
got
a
protocol
running
over
UDP
or
where
everything
needs
to
fit
in
one
packet,
you
should
be
very
concerned
about
keys
or
signatures
which
leads
to
the
next
one.
So,
as
Kenny
gave
a
presentation
for
me
yesterday
and
CFR
g
because
I
had
a
conflict,
I've
got
a
document
called
draft
hoffman
c
2pq,
which
is
classical
to
post
quantum,
where
it
has
nothing
to
do
with
what
will
be
chosen.
But
we
have
to
sort
of
get
an
idea
of
how
many
I'm
looking
for
people
to
help.
W
If
you
look
at
the
draft
you'll
notice,
it's
completely
full
of
holes
and
requests
for
for
input.
This
is
not
something
that's
at
all
baked.
We
need
a
lot
more
work
done
on
it,
we're
going
to
be
talking
to
other
people
in
both
in
the
physics
side
and
in
the
crypto
side,
but
in
this
room
we
are
going
to
want
to
know.
W
V
W
N
Feel
maker,
a
couple
of
points:
I
do
have
a
degree
in
nuclear
physics
and
I.
Don't
think
that
quantum
crypto
is
going
to
work,
because
the
sorry
quantum
key
exchange
will
work,
because
the
engineering
parameters
are
too
hard,
but
I
wanted
to.
N
One
point
is
that
we
do
have
Lamport
signatures
and
they
work,
and
it
might
well
be
worth
baking
in
Lamport
signature
roots
now,
because
that
is
one
thing
that
we
can
trust
and
we
don't
need
to
do
any
more
research,
but
going
back
to
your
original
bit
about
the
whole
sha-1
piece,
CAS
actually
started
moving
to
char
to
in
2008.
Some
of
them
did
yes
and
Russ.
Helped
me
push
one
of
them
to
do
just
that.
N
V
N
I
did
come
up
with
a
bunch
of
proposals
to
transmute
through
that
deployment
cycle
that
were
rejected
by
the
same
people
who
then,
when
they
could
see
the
writing
on
the
wall.
Come
back
that,
oh,
we
got
to
get
rid
of
this
right
now
and
we've
got
to
cancel
certs
that
are
already
been
issued,
so
I
mean
like
the
issues
that
were
you
reporting
to
on
cab
forum.
N
V
X
A
lot
of
this
is
the
application
space,
maybe
more
than
in
the
ITF
space,
but
I
think
it's
still
something
the
ITF
could
be
doing
is
knowing
that,
if
we're
creating
new
long
term
identifiers
and
those
can
be
correlated
with
things,
maybe
we
shouldn't
do
that.
Okay,
I,
don't
think
it's
just
identifier
stores,
also
the
plaintext
right
and
yes,.
X
V
V
V
U
Can
you
Nick
Sullivan
thanks
for
the
investment
advice
on
sergeant
II
base
crypto
but
I'm?
Something
to
point
out
that
actually
is
pointed
out
by
I
Sasha
new
base
crypto?
Is
that
the
key
sizes
don't
have
to
be
extra
large?
You
can.
Actually,
there
are
post
on
to
seed
schemes
in
which
you
trade,
a
lot
of
computation
with
small
key
sizes
for
quantum
resistive
quantum
Fourier
resistance
right.
V
I
think
that's
a
great
point
and
and
I
promise
to
not
put
any
equations
on
the
slides,
so
I
didn't
get
into
that
level
of
detail
I'm.
But
my
comment
about
so
Geneva's
crypto
was
mostly
about
the
novelty
of
it,
the
newness
of
it.
It
really
isn't
in
any
sense,
battle-tested
still
very
new
technology,
Tamiya.
Y
Y
V
Z
AB
A
AB
D
AB
We
also
could
you
go
back,
please.
We
also
applied
a
draft
here
or
supply
draft
here,
where
you
can
read
some
details.
Welcome
later
to
that
next
slide,
please.
The
problems
we
are
facing,
I
think
is
quite
clear.
Unfortunately,
there's
way
to
less
proofs
in
Internet
and
really
I
think
this
is
a
big
issue,
because
the
network
could
help
a
lot
for
civil
rights.
AB
But
if
it's
still
like
this,
then
we
need
to
improve
there
and
what
we're
actually
watching,
where
it's
going
to
encrypt
offer
users
is
either
it's
going
to
propriety
platforms
which
build
some
system
which
works
quite
well,
but
it's
closed
source,
no
one
else,
really.
What's
behind
the
curtain
and
on
the
other
side,
we
have
open
standards
which
are
yes,
basically
for
email,
some
others
and
which
are
quite
hard
to
use.
AB
AB
AB
So
our
idea
is
that
we
start
to
do
everything,
peer-to-peer
and
not
server
based-
and
this
is
the
core
of
the
PAP
idea,
having
a
peer
to
the
implementation
of
everything,
and
we
also
want
to
add
things
we
are
already
working
on
it,
which
is
hiding
metadata,
supplying
absolute
anonymity
and
anonymity
into
our
implementations
and
into
other
implementations.
So
this
means
extending
formats
extending
protocol
sense
of
and
one
really
important
thing,
which
is
totally
underestimated.
In
my
view,
is
we
we
never
managed
to
find
a
real
stand
of
Identity
Management.
AB
So
we
have
very
different
ways
to
manage
identity
and
at
least
we
need
a
mapping
or
we
can
even
define
what
sin
immunity
means
in
a
technical
way.
So,
yes,
so
what
we
want
to
achieve
is
a
common
concept
to
exchange
keys
and
signal
trust
until
the
user,
the
users
PAP,
is
a
very
user
centric
project.
We
compare
that
with
what
with
web
browser
guys.
That
means
we
end
up
with
something
like
a
traffic
light
semantics
next
slide.
Okay,
so
what
are
those
guys
already
doing?
AB
We
already
have
an
implementation
of
automated
key
management
locally
and
between
devices
that's
automatic
key
management.
We
mean
completely
peer-to-peer,
we
mean
generation,
revocation
of
keys.
We
call
a
key
reset,
which
is
a
little
complex
thing,
so
I
need
your
keys
because,
and
we
already
implemented
a
three-stage
trust
rating
concept.
We
implemented
as
many
open
PGP,
because
open
standards
highly
important
for
us,
we
want
to
be
as
comfortable
as
possible.
We
implemented
message
formats
which
have
a
little
more
privacy
properties.
AB
We
implemented
a
first
version
of
what
we
call
peace
in
which
is
synchronizing
of
keys
between
the
devices
one
user
is
owning.
That
means
I
have,
let's
say
a
pad,
and
a
telephone
and
the
laptop
and
I
want
to
share
the
same
keys
on
all
my
devices,
but
no
one
others
should
have
the
keys
of
obvious
reasons.
So
we
implemented
that
and
we
created
adapters.
AB
So
you
can
use
the
implementation,
because
all
what
we
are
doing
is
free
software
in
common
languages,
where
applications
are
built
next
slide,
please
what
we
also
did
is
we
just
implemented
applications
to
show
that
this
is
actually
working
and
you
can
just
download
and
try
it
out.
So
we
did
it
for
couple
of
platforms
and
what
is
still
missing
is
we
have
a
deal
with
the
new
I
think
it
was
first
time
in
history
that
new
actually
is
giving
out
property
and
instead
of
techadon
and
first
guys
they
did.
AB
It
is
us
we
actually
co-owning.
Now
we
have
ownership
on
unit
as
pet
foundation,
and
we
are
building
me
that
the
DHT
base
onion
implementation
based
on
the
subsystem
of
new
net,
which
is
coca
debt.
And
yes,
we
have
way
more
work
to
do
because
we
need
to
synchronize
more
things.
So,
for
example,
we
already
synchronize
keys
would
be
nice
to
also
have
to
trust
synchronized.
If
you
trust
something
on
one
device,
it
would
be
nice
that
it's
trusted
on
the
other
device,
for
obvious
reasons,
thanks
light.
AB
AB
AB
So
we
can,
you
guys,
help
us
I
hope
or
we
can
be
a
new
member
of
the
family,
which
I
hope
more.
This
is,
of
course,
there
is
a
draft
for
man
based
message
formats,
which
is
from
the
ITF,
and
we
already
implemented
that
for
being
compatible
in
our
engine.
But
unfortunately,
this
draft
has
nature
issues
in
our
concept.
And,
yes,
let's
say
we
are
not
very
happy
on
it.
AB
That
is
the
reason
why
we
will
do
another
one
and
let's
work
on
that,
and
we
have
the
hope
when
we
design
at
work
protocol
families
which
could
be
helpful
not
only
for
pepper
from
many
protocols
like
doing
a
peer-to-peer,
kissing
implementation.
We
have
the
hope
that
we
can
make
a
standard
that
this
really
works
over
platforms
and
protocols.
AB
Yes,
also
the
mappings.
You
know
we
are
using
email
and
travel
for
in
inland
communication
of
things
like
that
kissing,
and
so
we
want
to
standardize
that
so
and
we
don't
get
into
the
mess
that
the
different
implementations
will
not
be
comfortable
and
there's
even
things
like
UI
schemes
we
design,
because
internally,
we
are
using
your
eyes
for
addresses
and
next
slide.
Please
there's
also
a
lots
of
future
topics.
We
have
hope
to
find
people
being
interested
in
the
ITF,
which
is
identity.
AB
Mapping
kalenna
mapping,
as
already
mentioned,
and
also
what
we
call
key
many
key
mapping,
which
is
we
have
text
message
transports.
We
call
them
transfer
to
email,
isn't
transport
for
pep,
which
means
we
have
such
things
where
a
PKI
is
just
a
huge
and
you
cannot
use
it
on
such
small
things.
Think
of
SMS
ism,
a
primitive
example,
and
so
we
have
automaticall
mapping
onto
symmetrical
temporary
keys
and,
of
course,
we're
also
working
on
non
mind
message
formats
I
already
man
I,
mentioned
Java
I,
don't
know
how
much
they
love
it
next
slide.
AB
So
what
is
our
hope
to
find
here?
Yes,
we
really
want
interoperability
and
and
I
think
this
is
the
right
place
to
ask
for
it
and,
of
course
we
need
lots
of
critics.
We
need
your
heart
critics
because
we
are
just
doing
it
on
our
own
and
just
put
an
implementation
on
the
table
here,
and
it
would
be
nice
to
hear
the
opinion
of
experts
in
these
fields
where
we
are
right
and
may
be
wrong
right,
and
what
we
also
hope
is
that
foundation
is
way
too
small
to
offer
a
management
of
of
open
standards.
AB
AB
AB
T
Zone
R
has
the
pepper
nation
ever
taken
a
look
at
the
work
in
the
Utah
working
group
with
regard
to
email,
security
reporting
and
UI
/
security
improvements.
As
far
as
I
know,
pep
is
implementing
a
kind
of
overlay
for
popular
clients
based
on
either
s1
or
jpg.
The
PGP
Utah
meets
later
on.
Please
join
so
I
believe
he's
suggesting
that
that
would
be
one
place
that
meets
the
requirement
in
your
last
slide.
Last
bullet.
Okay,
that's!
Where
would
this
take
place?
Utah
appears
to
be
one
set
of
place.
Well,
that's
great
input.
Thank
you.
T
AB
We'll
do
that?
Okay,
so
because
I
only
have
four
quarter
of
an
hour.
I
know
the
approach
is
way
too
complex
to
explain
it
in
a
quarter
of
an
hour.
I
would
love
to
end
my
people.
We
do
a
bob
off
in
room
Terrica
this
evening
and
if
you
have
time
feeling
lighted
and
this
for
sure
will
end
up
in
a
bar
I
can
guarantee
event.
Thank
you.
AC
Hi
pal
voters
so
I'm
not
exactly
sure
what
you're
proposing
are
you
is
this
an
effort
to
strengthen
a
whole
lot
of
protocols
within
the
IETF
to
do
better
security,
or
are
you
suggesting
that
this
is
a
whole
new,
federated
area
that
we're
building,
or
are
you
some
new
centralized
phone
book
that
has
all
these
privacy
and
anonymity
features
inside
like
that
I'm?
Really
confused?
Okay?
First,
we
never
centralized.
AB
We
do
only
peer-to-peer
and
there's
no
central
infrastructure
and
the
pet
foundation
is
managing
nothing.
Okay,
so
then
hi
I'm
Edward
Snowden,
here's
my
identity,
yes,
which
is
possible
because
we
don't
manage
identities.
I
talked
about
mapping
of
identity
management
of
other
people
which
they
do
in
different
ways
into
each
other.
So
there
is
interoperability.
It's
not
meant
that
pet
foundation
will
not
provide
a
namespace
where
you
can
register.
AD
Okay,
so
mutton
comes
and
that
actually
helped
a
little
bit.
I
read
your
draft
I've
looked
at
your
website
in
some
detail
and
I
have
to
confess
that
neither
of
those
things
helped
me
that
the
slightest
bit
understand
what
you're
trying
to
do.
I
saw
some
statements
about
grand
principles
and
I
saw
some
code,
but
I
didn't
see
anything
connecting
the
two
of
them.
AD
But
you
need
to
ask
those
people
for
help,
because
you're
actually
asking
for
a
number
of
different
things.
Yes,
and
what
we.
What
the
people
here
would
like
to
see
is
how
the
what
your
goals
are
and
very
in
a
very
concrete
way,
how
those
goals
fit
together.
If
you
have
an
architecture,
what
that
architecture
looks
like
and
how
the
various
pieces
fit
together
and
where
you
have
identified
deficiencies
in
the
individual
protocols
that
you're
using
what
those
deficiencies
are
and
how
you
propose,
they
would
be
fixed,
and
this
is
a
massive
project.
AD
There's
working
implementation
already,
which
we
can
download
and
try
out
so
my
caution
that
there's
there's
a
very
big
difference
between
implementing
some
code
and
shipping,
some
code
and
actually
getting
to
the
point
where
you
have
interoperability
and
that's
a
long,
hard
battle
and
I
roundly
congratulate
you
for
even
attempting
that.
Thank
you.
Well,.
N
To
the
home,
Baker
yeah
I
I
been
following
this
for
quite
a
while
now
and
I
like
it
and
I
just
like
to
point
out
at
the
moment,
there
are
a
lot
of
usable,
secure,
end-to-end
messaging
mechanisms
around
you
know,
we've
got
signal
whatsapp's
and
so
on.
There
are
two
problems
of
that:
one
is
that
they're
all
single
silo
models.
N
The
implementations
can
be
audited
and
the
protocols
have
to
be
designed
in
such
a
way
that
when
you
are
dealing
with
a
messed
up
version
of
the
a
coerced
version
of
the
application
that
tries
to
talk
to
a
non
coerced,
then
it's
got
to
become
apparent
and
one
of
the
things
that
I
had
when
I
talked
to
somebody
very
senior
in
one
of
the
agencies.
Formally
his
advice
was
Nobis.
N
Nobody,
but
us
in
that.
The
thing
is
that
they're
very
willing
to
mess
around
with
crypto
and
applications
or
whatever,
but
they
are
very
adverse
to
having
their
messing
up
being
detected
sure,
and
so
that's
the
thing
that
we've
got
to
work
for
whether
this
is
I've
got
problems
with
this
particular
implementation.
I
think
that
maybe
we
need
to
go
to
a
consensus
algorithm
that
allows
us
to
centralize
without
being
centralized,
but
it
could
work
Thanks.
C
C
AE
AE
When
Google
found
that
Symantec
misbehaves
the
company
suggested
to
limit
certificates
issued
by
Symantec,
it
suggested
to
reduce
the
accepted
validity
period,
the
to
get
the
distrust
to
current
us,
that
the
cemetery
shield
certificates
and
so
on
and
so
forth,
and
even
to
remove
their
recognition
of
external
validation
status
for
Symantec
issued
certificates
next
slide,
please
now,
when
application
wants
to
limit
trust,
it
has
in
fact
thought
a
options.
Well,
application
can
specify
which
processes
are
valid
for
in
its
context,
it
can
turn
off
Sorrell
and
SSP
validation,
and
we
have
problem.
AE
What
shall
we
do
when
SSP
is
not
available,
and
in
case
we
have
flex
availa?
We
need
flexible
limitation
in,
for
example,
when
one
of
say
is
too
big
to
fail
week
and
we
have
to
have
poured
them
next
slide.
Please
I
propose
a
solution,
certificate
limitation,
name,
the
certificate
limitation
profile,
it's
a
list
that
can
be
distributed
with
the
application
or
by
the
application
as
a
part
of
application
bundle.
AE
If
it
should
have
sailor-like
syntax,
it
should
be
cryptography
designed,
and
if
this
proposal
finds
interest
and
implemented,
the
users
application
can
use
code
base
instead
of
application
level
hot
coding.
This
will
allow
to
avoid
some
some
common
mistakes
in
validation
next
slide,
please.
So
what
can
be
stored
as
examples
of
such
limitations?
At
first,
we
should
say
whether
the
limitations
are
applied
to
certificate
itself
or
its
descendants
on
the
chain
of
trust,
and,
for
example,
we
can
say
we
should
not
trust
if
agates
issued
after
this
other
date,
we
should
not
test
their
certificates.
AE
Matching
the
profile
after
the
concrete
date,
we
can
limit,
develop,
definitive
certificates
that
we
can
say
which
expire
and
extension
should
be
ignored.
We
can
say
which
they,
which
extensions,
should
be
quite
difficult,
for
example,
for
replica
it's.
The
common
case
is
silica
transparency
related
extensions.
AE
AE
Implies
changes
to
the
verification
process
after
building
change
chain
of
trust
we
go
through
it
and
apply
limitation
to
the
matching
sets
and,
if
specified
in
the
profile
we
apply
them
to
is
the
sandals.
If
we
find
that
the
certificate
should
not
be
trust
that
we
stopped
verifying
and
return
that
it's
entrusted
next
slide,
please,
and
so,
instead
of
hard-coding
many
a
limitation
at
the
application
level,
we
can
limit
very
small
set
of
check
that
we
have
cell
P
in
use
that
the
selfie
you
operate
is
a
fraction
of
that.
AE
C
AD
I
don't
know
what
this
is,
but
having
said
that,
why
exactly
at
the
mic,
I
I
certainly
haven't
read
this
but
you're
describing
a
number
of
things
that
has
a
significant
amount
of
overlap
with
a
number
of
his
existing
deployed
mechanisms
like
public
key
pinning
and
we
implement
a
thing
called
one
Sorel
must
have
been
a
proposal
to
use
recursive
bloom
filters
in
combination
with
the
certificate
transparency
log,
to
have
an
exhausting
exhaustive
list
of
certificate
revocations.
Why
would
this
be
superior?
Well,
their
certificate.
AE
S
Yeah,
yes,
so
the
there
are
application,
then
Richard
Barnes.
There
are
application
vendors
doing
this
today,
so
a
chrome
google
for
chrome
produces
here,
l
sets
which
are
basically
largely
overlapping,
with
kind
of
the
limited
subset
of
the
functionality
you
described
and
mozilla
produces
a
thing
called
one
crl,
as
Mark
mentioned
for
Firefox,
and
so
the
question
is
those
are
proprietary
things
about
there
between
a
vendor
and
its
applications.
So
I'm
curious,
like
what's
the
value
of
standardizing
this
and
adding
as
all
this
additional
functionality.
S
AE
K
This
is
Daniel
Kang
Gilmore,
with
SEIU
I'm,
also
work
on
the
Debian
project.
I,
don't
know
if
you're
aware
of
the
p11
Glu
project,
which
has
a
storing
trust
policy
specification,
they've
been
working
on
that
fits.
What
you're
doing
I
think
almost
exactly
and
the
use
case
Richard
is
that
there
are
a
bunch
of
different
things
that
actually
check
certificates
and
the
rules
are
subtle,
and
so,
if
you
can
actually
build
this
particular
API,
then
you
can
encourage
people
to
not
rewrite
their
own.
So
it
is
it
days.
K
R
Rich
Sol's,
I
guess
open
a
cell
affiliation
yeah,
it's
great
that
two
or
three
programs
on
the
Internet
can
do
really
fast
and
really
precise.
Checking
of
revocation
but
I'd
like
to
see
this
extensible
to
all
sorts
of
things.
So
not
just
the
browsers,
but
there
are
many
web
PKI
programs
out
there
so
I
support.
This
would
be
interesting
to
come
up
with
a
standard
data
format.
I've
had
some
discussions
with
some
of
the
Linux
vendors
about
doing
that.
So
yeah
I
encourage
this.
AD
S
So
there's
Richard
Barnes
again
so
that
thinks
TKG
for
the
clarification
there
I
think
yeah
I
can
see.
There
could
be
some
value
to
this
I
think
at
least
a
soft
prerequisite
for
this
work
is
going
to
be
well.
A
prerequisite
for
the
work
to
be
successful
is
going
to
be
some
notion
of
who
is
trusted
to
distribute
this
information
in
the
you
know,
vendor
to
its
application
space,
it's
kind
of
clear,
because
you
trust
Google
for
Chrome.
You
trust
in
Mozilla
Firefox.
S
D
Eric
rajala
for
for
someone
else,
one
of
those
projects,
the
there
actually
are
two
problems
to
solve.
One
problem
is
the
policy
problem
of
indicating,
like
which
anchors
you
trust
and
which
intermedius
to
revoke,
and
the
other
was
to
address
the
problem
of
any
certificates
and
the
and
the
problem.
There
is
data
compression,
because
the
number
of
road
initiative
is
far
too
large
to
actually
send
over
in
any
format.
Books
like
this
at
all,
so
so
just
have
an
entire
problem.
You
gotta
solve
that
problem
as
well.
AE
D
C
X
Hi
Steven
Farrell,
so
there
was
a
bit
of
an
out
force
yesterday
in
today's
session
and
hopefully
we
killed
off
one
one
more
bad
idea.
So
what
I?
Without
getting
into
the
details
of
that
which
I
don't
think
are
right
here
and
the
TLS
working
group
are
continuing
to
beat
them
more
bad
idea:
ferries,
I
guess!
X
If
the
TLS
work,
we
went
crazy,
I
think
that
wouldn't
be
in
their
Charter
and
what
they
would
need
to
rechart
her
to
do
the
crazy
thing
and
also
I
think
he
heard
from
Lars
and
from
Kenny
that
clearly
such
such
changes,
TLS
would
have
big
impacts
outside
one
working
group
and
outside
the
security
area
in
the
other
parts
of
the
ITF.
So
I'm
just
wondering
what
you
guys
think
about
it.
X
C
C
Hey,
that's
one
thought
that
I
had
after
listening
to
some
of
the
conversations-
and
you
know
so
you
know
cueing
off
of
other
people's
ideas
and
thoughts.
You
know
maybe
there's
an
opportunity
for
something
data
center
specific
and
then
you
could
design
just
what
you
need
and
it
would
interfere
with
TLS
or
you
know
our
internet
level.
Applications
for
encryption
I
know
this.
One
thought
I
had
yeah.
X
D
F
C
D
X
D
J
Sorry,
probably
less
interesting,
but
still
back
it
back
to
Kenneth's
presentation.
There
was
a
quantum
random
Oracle
mentioned
in
CFCF
og
and
you
guys
need
a
theoretical
models
to
prove
security.
So
I
wanted
to
ask
probably
your
personal
opinion
on
whether
these
models
exist,
I,
reliable
and
something
that
that
NIST
can
base
and
the
evaluation
on
I.
V
Three
yeses
I
was
stable,
reliable,
well.
The
quantum
random
welcome
well
was
something
we're
still
getting
our
heads
around
there's
a
few
experts
in
the
world
who
know
how
to
work
with
it,
but
the
group
of
people
who
can
is
growing
your
time.
It's
basically
an
extension
of
the
random
Oracle
model,
which
is
some
kind
of
abstraction.