►
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
Oh
yeah,
so
to
start
off
with
some
use
cases,
yeah
we've
already
gone
over
some
of
these,
but
just
to
re-summarize,
I
guess
so.
For
ipfs,
some
of
the
main
use
cases
are
for
lookup
privacy.
So
obviously,
if
you
want
to
look
up
some
sort
of
information
that
may
be,
I
don't
know
sensitive,
you
know,
maybe
you
don't
want
people
to
know
you're,
looking
it
up
or
it's
like
politically
banned
or
whatever.
Basically,
there's
right.
A
A
So
potentially
things
like
transaction
submission
for
a
blockchain,
so
there's
sort
of
some
research.
That's
been
done
on
this
previously
with
like
dandelion
but
yeah.
Potentially,
we
could
use
a
similar
idea
for
yeah
blockchains
as
well,
just
because
a
lot
of
blockchains
now
do
use
the
p2p
like
yeah,
youtube
and
polkadot
and
falcon,
and
then
there's
also
the
kind
of
generic
case
that
people
usually
talk
about,
which
is
like
a
p2p
chat,
app
yeah.
A
But
then,
for
I
guess
this
presentation
I'm
going
to
focus
mostly
on
lookup
privacy
or
I
think,
as
yeah
it
was
previously
referred
to
as
like
reader
privacy,
so
yeah,
essentially
preserving
privacy
for
someone
looking
up
something
in
the
system:
okay,
so
yeah,
so
brief
overview
of
unwriting.
So
this
is
what
torque
uses,
as
you
guys,
probably
all
know.
So
it's
basically
like
a
layered
encryption
approach.
So
basically
you
have
a
circuit
consisting
of
generally
three
nodes
between
the
source
of
traffic
and
then
the
destination
of
traffic.
A
And
then
you
have
the
public
keys
of
all
the
the
nodes
in
between
and
then
you
kind
of
wrap
the
message
in
layers
of
encryption,
so
starting
with,
firstly,
encrypting
with
the
destination,
then
the
exit,
node
and
the
middle
then
the
entry
and
then,
as
you
kind
of
move,
the
message
through
the
system,
each
layer
of
encryption
gets
unwrapped.
So
basically
this
provides
privacy
for
the
traffic
originator
as
as
none
of
the
nodes
in
the
system.
Apart
from
the
entry
node
know
who
it
is
so
this
is,
I
don't
know
pretty.
A
I
don't
know
basic
kind
of
like
low-hanging
fruit.
In
some
ways
of
providing
originator
privacy
yeah,
so
I
guess
for
this-
you
have
to
be
sure
that
the
entry
and
exit
nodes
don't
collude.
So
that's
kind
of
one
of
the
open
questions.
If
we
do
this
in
a
in
a
p2p
way,
see
I'll
talk
about
that
a
bit
more
later
and
then
mixnets
so
make
sense
kind
of
build
a
lot
on
top
of
this
idea
and
add
a
lot
of
other
techniques
for
providing
stronger
privacy,
so
yeah
so
with
onion
roading.
A
As
I
kind
of
mentioned,
you
need
to
be
sure
that
the
entry
exit
nodes
don't
collude,
and
it's
also
since
it's
still
just
like
one
message
being
relayed
essentially
through.
Like
a
few
other
nodes,
you
can
still
you're
still
vulnerable
to
things
like
time,
analysis
and
like
traffic
analysis
and
that
kind
of
thing,
so
a
mixnet
adds
basically
more
techniques
to
prevent
against
that.
A
So
it
uses
an
eroding
as
well,
but
it
also
uses
other
techniques
like
combining
messages
together
into
like
a
fixed
size,
bundle
holding
messages
potentially
for
certain
amounts
of
time.
To
kind
of
I
don't
confuse
the
the
traffic
flow
and
also
uses
decoy
traffic,
so
yeah
just
fake
traffic
to
add
more
confusion.
I
guess
to
the
mix
so
yeah,
as
you
can
see
like
in
the
image
it's
kind
of
each
mixed.
Node
takes
bunch
of
messages,
bundle
them
together,
re-shuffles
them
like
sends
them
to
the
next
one.
A
So
then,
by
the
end,
you
basically
have
broken
the
link
completely
so
yeah,
so
some
existing
and
non-empty
networks,
so
yeah,
like
I
mentioned,
there's
tor.
So
this
uses
just
on
your
routing
essentially
and
its
main
purpose
is
for
anonymizing
like
internet
access,
so
it's
not
yeah.
It
doesn't
really
apply
necessarily
to
like
ipfs
or
lib
p2p,
as
it's
not
really
peer-to-peer.
It's
more
like
you,
I
don't
know,
enter
the
network.
A
You
find
your
circuit
from
the
directory
nodes
and
then
access
the
internet
that
way,
but
yeah,
it's
well
known,
it's
popular,
so
yeah
and
then
there's
i2p
or
the
invisible
internet
protocol
which
is
different
from
tor
and
that
it's
more
it's
p2p.
It's
kind
of
a
closed
network.
In
that
you
only
access
things
inside
the
network.
Like
you,
don't
really
access
on
the
outside
internet
and
it
uses
garlic
roading
in
addition
to
any
routing.
A
So
it
does
some
message
bundling
as
well,
but
yeah
itp
is
like
not
really
as
much
used
as
tor.
Unfortunately
so
yeah
and
there's
not
been
really
a
lot
of
research
on
it
either
so
yeah,
it's
it's
kind
of
there
and
then
there's
dandelion
or
daylight
plus,
which
is
designed
specifically
for
transaction
submission
privacy.
So
it
was
used.
I
don't
know
if
it's
actually
using
bitcoin,
but
it's
definitely
used
in
monero
and
it's
yeah.
A
So
basically
it's
just
it's
a
message
or
it's
a
method
for
when
you
send
the
transaction
to
the
network,
basically
gossiping
it
in
a
certain
way
that
preserves
privacy
so
yeah,
you
could
kind
of
extend
it,
maybe
to
some
other
gossip
protocols,
but
it
was
designed
specifically
for
blockchains
and
then
yeah
and
then
a
few
modern
mixednet
projects
so
yeah,
I
think
yeah,
so
there's
a
name
which
is
an
incentivized
mix.
Actually
all
of
these
are
incentivized
but
yeah.
So
nim
is
an
incentivized
mixnet.
A
It
uses
the
syncs
packet
format
and
then
it
the
mix,
delays
the
packets
instead
of
waiting
for
a
specific
threshold
and
then
bundling
them
together.
So
the
point
or
what
they
say,
the
purpose
of
this
is
is
to
basically
improve
the
latency
so
that
you
don't
have
to
wait
for,
like
a
certain
amount
of
traffic
to
reach,
to
reach
the
mixed
node.
You
can
just
kind
of
randomly
delay
it
instead
and
then
there
is
cmix.
So
this
is
part
of
the
xx
network.
So
this
is
also
incentivized.
A
A
So
then,
basically,
how
can
we
apply
this
to
the
p2p
and
ipfs?
So
one
method
is
unenrolling.
So
generally
three
hop
underwriting
is
like
the
traditional
kind
of
underwriting
that's
used,
and
then
you
could
use
this
for
providing
lookup
privacy
so,
for
example,
like
dht
lookups
or
looking
up
cids
and
ipfs.
A
So
it's
sort
of
simple
low
hanging
fruit.
In
that
regard,
you
could
do
like
a
very
kind
of
basic
design
where
you
kind
of
randomly
pick.
I
don't
know
three
modes
and
then
make
a
circle
to
them,
but
then
you
could
get
range
it
from
kind
of
simple
that
maybe
doesn't
have
that
good
of
security
to
like
a
really
a
more
complex
system
that
has
like
maybe
distinct
node
types
more
like
advanced
methods
of
circuit
selection,
maybe
yeah
having
like
specific
nodes.
A
So
this
would
only
provide
look
at
privacy
doing
the
other
way
around
so
getting
a
node
to
act
as
like
a
hidden
service,
essentially
or
basically
to
provide
anonymity
for,
like
the
receiver's
message
is
much
harder
like
torah
has
uses
like
rendezvous
points
like
a
whole
lot
of
other
yeah
implementations
to
achieve
this
so
yeah,
so
that's
yeah,
unfortunately
not
covered
by
this
and
then
yeah.
And
then
I
mix
that,
like
design,
protect
more
against
traffic
analysis
and
that
kind
of
thing
so
yeah,
it
could
kind
of
range.
A
From
very
simple
and
somewhat
naive,
I
guess
to
very
complex
depending
like
how
how
much
trade-off
we
want
to
have
in
terms
of
like
latency
and
that
kind
of
thing
so
yeah,
there's
yeah,
so
some
open
questions
that
we've
discovered
in
our
group
relating
to
this.
So
one
of
the
main
ones
that
I've
kind
of
mentioned
already
is
how
our
circuit
selected.
A
So
this
seems
to
be
like
really
carefully
considered,
because,
basically,
if
entry
and
your
entry
and
exit
nodes
collude,
then
you're
kind
of
screwed,
because
they
can
just
talk
to
each
other
and
then
know
where
the
message
is
going,
where
it's
coming
from
so
yeah.
We
need
to
very
carefully
consider
this.
I
would
say,
and
then
do
we
require
a
structured
network
for
similar
security
guarantees
to
tour.
A
So
do
we
need
nodes
to
specifically
be
like
entry
relay
and
exit,
or
can
we
have
like
more
of
a
p2p
method
where
any
node
can
be
any
role
and
then
do
we
need
some
nodes
to
act
as
directory
authorities
and
basically
have
a
full
view
of
the
network
like?
Is
it
enough
to
gather
the
circuit
from
just
like
your
peers
or
like
whatever
your
peers?
Tell
you
or
do
you
have
to
like?
A
Have
some
notes
be
kind
of
authorities,
and
then,
if
we
do
that,
then
how
do
you
kind
of
make
it
not
just
completely
centralized
and
then
yeah
then
also?
How
do
we
take
into
account
network
churn?
So
if
you
have
like
a
circuit
and
then
one
of
the
nodes
just
leaves
or
whatever,
then
how
do
you
deal
with
that?
So
a
couple
potential
solutions
are
just
fallback
circuits
or
having
like
multi-path
circuits.
A
A
It
has
like
it
didn't
fully
end
up
working
because
other
components
still
need
to
be
modified
and
I'm
going
to
talk
about
a
little
bit
more
about
that
later
as
well,
and
then
how
is
latency
going
to
be
affected
specifically
for
dht
lookups
and
then
also
do
we
make
privacy
often
or
mandatory
on
a
specific
network,
so
yeah,
that's
kind
of
a
big
one.
I
would
say
because
yeah
kind
of
relating
to
other
ones
as
well
like
we
want
a
bigger
anonymity
set.
A
Obviously,
but
people
might
not
want
to
do
the
might
might
not
want
to
have
latency
trade-offs.
They
might
want
to
yeah
not
really
have
to
deal
with
that
so
design
kind
of
concerns
there.
And
then
how
do
we
deal
with
the
bootstrap
problem?
So
since
we
need
a
certain
amount
of
nodes
on
the
network
to
have
like
a
good
anonymity
set
and
actually
like
get,
I
don't
know
an
underwriting
or
mixed
type
design
to
work.
A
How
do
we
kind
of
deal
with
that
when
the
network
is
small
and
then
how
can
a
network
so
say
we
do
implement
this
into
the
pdp.
So
how
can
a
a
network
like
isolated
that
has
these
privacy
features
built
in
utilize,
like
the
anonymity
sets
of
other
networks?
That
also
have
this
built
in?
So,
as
we
all
know
like
more
nodes
running,
this
would
be
much
better,
obviously
so
kind
of
do.
A
They
all
use
a
blockchain
as
a
form
of
incentivization
to
get
the
nodes
to
actually
perform
the
mixing
correctly
so
say
yes,
let's
say
we
do
kind
of
go
with
a
design
like
do.
We
need
the
nodes
to
be
incentivized
to
actually
mix
or
relay
correctly
and
yeah
like
if
we
don't
like,
for
example,
like
with
tor
like
it's
all
volunteer
based
like?
Are
there
problems
there
that
could
arise
like
could
people
just
say
like?
A
I
don't
want
to
do
it
and
then
just
like
disable
it
within
their
note
or
something
like
that
so
yeah,
another
open
question
there
yeah
and
then
potential
modifications
to
the
pdp,
so
yeah
sort
of
touch
on
this,
but
a
couple
areas
that
I
think
pretty
obviously
would
need
to
be
modified
if
we
end
up
doing
this
so
there's
autonat,
which
so
a
node
basically
asks
other
nodes
to
dial
its
presumed
public
address.
So
this
obviously
like
leaks
its
public
address
so
kind
of
to
yeah.
A
So
to
deal
with
this,
we
would
have
to
modify
autonod
in
some
way
like
we
could
have
it.
Maybe
only
ask
specific
nodes
that
it's
going
to
already
leak
its
address
to
already
to
do
this,
but
yeah
that
may
not
work
and
there's
identify
protocol.
So
this
also
this
happens
when
two
laptops
connect
and
then
it
shares
like
the
nodes
publicly
and
it's
listening
addresses,
which
also
leaks
its
public
address
so
yeah
this
would
have
to
be
modified
as
well.
A
So
basically
like
the
more
the
more
you
use
like
mix
net
type
techniques,
the
greater
impact
you're
gonna
have
on
the
latency,
so
yeah,
first
of
all,
like
obviously
a
multi-path
circuit
is
gonna
increase
the
latency
having
stuff
like
message.
Delays
is
gonna,
obviously
increase
latency
as
well.
So
we
have
to
kind
of
consider
like
how
much
trade-off
do
we
wanna
have
there
and
same
with
bandwidth
for
just
plain
underrating
bandwidth
isn't
really
affected
as
much
but
for
stuff.
A
Like
other
mixnet
designs
like
if
you
have
like
bundling
that
kind
of
thing,
it
definitely
will
affect
the
bandwidth,
there's
circuit
setup
time
so
like
how
long
does
it
actually
take
you
to
set
up
the
circuit
like?
How
long
will
it
take
you
to
find
the
right
nodes,
establish
the
connections
and
exchange
public
keys,
encryption
time,
that
kind
of
thing
and
then
obviously
the
final
thing
is
like
the
actual
privacy
or
like
the
anonymous
set.
As
you
know,
bigger
is
better
and
yeah
more
notes.
A
There
is
always
better
so
yeah.
These
are
kind
of
some
of
the
yeah
main
metrics
to
consider
cool.
I
think
that's,
basically
it
so
yeah.
Thank
you
for
listening
everyone.
If
yeah,
I
would
love
to
discuss
this
more
with
everyone.
So
if
you
have
any
ideas
or
thoughts
or
whatever
about
the
open
questions
or
just
any
feedback,
that
would
be
much
appreciated,
so
yeah
feel
free
to
find
me
or
contact
me
my
email
yeah.
Thank
you.
Everyone.
B
A
Yeah
yeah,
that's
a
good
question
yeah,
so
with
nim.
I
know
that
they
have
like
specific
mix
nodes,
so
these
are
like
different
from,
like
the
say,
like
the
client,
that's
actually
using
the
network.
So
as
a
client,
you
want
to
have
anonymity,
but
then,
as
like
a
mixed
operator,
you,
I
guess,
don't
care
about
it
as
much.
So
you
essentially
from
what
I
know
you
don't
have
really
as
much
anonymity
there.
It
focuses
more
mostly
on,
like
the
client
part
yeah
and
then
for
yeah
for
hopper.
B
A
B
A
A
Yeah
yeah,
that's
a
really
good
question
yeah.
I
don't
think
we
have
like
any
specific
measurements
yet,
but
I
guess
for
unknown
if
you're
gonna
have
like
three
hops,
so
that's
like
three
extra
nodes
and
basically
would
be
like
on
average
like
latency
times,
four
or
so
so,
that's
kind
of
like
I
guess
the
minimum
latency
kind
of
overhead
yeah.
So
obviously,
if
we
added
more
mixed
stuff,
it
would
be
more
than
that.
So.
A
Okay,
yeah,
you
mean
for
like
dht
lookup,
specifically
because
it's
going
to
go
through
like
multiple
dish,
you
notes
yeah
yeah,
that's
a
good
question.
I
guess
I
guess
the
first
thing
that
comes
to
mind
is
kind
of
what
I
asked
you
previously,
which
is
like
doing
like
the
recursive
lookup
through
the
dht
instead
of
like
the
iterative.
So
I
believe
there
is
some
issue
with
doing
the
recursive
lookup,
but
essentially
like.
A
If
you
did
that,
then
each
node
would
it
would
like
yeah
forward
the
traffic
from
the
previous
node
and
then
not
know
like
which
node
was
the
originator,
so
that
is
kind
of
a
similar
idea.
I
guess
that
could
be
implemented
yeah,
so
it
might
not
yeah.
It
wouldn't
be
underwriting.
Basically,
if
you
added,
if
you
did
recursive
lookup
instead,
I
think.
C
B
And
then
do
the
whole
query
for
you,
and
that
has
less
of
a
amplification
worry,
because
it
knows
that
you
did
the
effort
of
constructing
the
circuit
to
establish
that.
So,
if
doing
multiple
pops
of
work
for
your
one
request
is
lesser
you've
actually
done.
It
knows
that
you've
done
multiple
circuit,
setups.