►
From YouTube: PoRep Day 2020 // What are PoReps? - Irene Giacomelli
Description
For more information on Filecoin
- visit the project website: Filecoin https://filecoin.io/
- or follow Filecoin on Twitter: https://twitter.com/Filecoin
Get Filecoin community news and announcements in your inbox, monthly: http://eepurl.com/gbfn1n
For more information on IPFS
- visit the project website: https://ipfs.io
- or follow IPFS on Twitter: https://twitter.com/IPFS
Sign up to get IPFS news, including releases, ecosystem updates, and community announcements in your inbox, each Tuesday: http://eepurl.com/gL2Pi5
A
Thanks
one
for
the
four
this
morning,
traduction
and
this
we
are
going
to
have
three
talks
this
morning.
This
is
the
first
one,
so
I'm
going
to
start
from.
Basically,
they
I
really
try
to
construct
the
definition
of
power
up
that
is
going
to
be
than
the
main
topic
for
the
nest.
2
toxin
and
I
would
like
to
construct
it
parting
from
starting
from
the
proof
of
space.
A
So
basically
out
of
the
talker
almost
would
be
more
onto
4
space,
because
I
started
to
understand
this
definition
since
prefer
replication
construct
on
top
of
proof
of
space,
proof
of
space.
I'll
say
a
bit
more
about
it
in
a
moment,
but
it's
just
the
intuition.
We
want
a
replacement
for
proof
of
work.
A
So,
as
I
said
before
the
proof
of
space,
the
motivation
was
to
have
an
alternative
of
proof
of
work.
You
can
use
the
application
of
a
proof
of
space
that
are
the
same
that
we
have
for
proof
of
work,
so
it
can
be
spam
prevention
or
consensus
protocols,
a
was
the
introduced
introduced
and
then,
quite
recently,
a
crippled
2015
by
Stefan
and
we'll
roasting,
maja
and
and
and
others,
and
basically
it's
interacting.
It's
a
protocol
between
a
poverty,
verifier
and
the
high
level
goal.
A
A
We
are
basically
the
the
goal
of
the
linearization
phase
is
for
the
proven
the
verifier
to
produce
this
string
as
it's
the
filler
from
the
space.
So
it
has
to
be
longer
like
if
my
parameter
for
measuring
the
space,
the
storage,
we
call
it
and
he
has
to
be
an
anthology
units
and
units
and
instead
of
very
far
just
as
more
advise
a
very
short
string.
This
communication
has
to
be
efficient,
I'll
go
back
on
this
in
a
second,
and
when
we
are
done
with
this,
we
can
have
this
execution
phase
to
repeat
every
time.
A
A
So
if
you
think
about
this
in
this
is
time
what
is
going
to
up
is
that
we
have
this
longer
say
more
than
a
second,
this
more
heavier
part
that
is
in
sterilization
and
then
through
time.
I
repeat,
this
is
a
Kishin
I,
mean
program,
verify
repeat,
pieces,
equation,
phase
and
and
troll
this
time.
I
can,
if
the
verifies
saying
good
proof,
so
you
can
say
yes
to
all
the
proofs.
A
We
are
secure
that
the
prover
is
still
storing
us
through
all
this
time,
even
between
choice
occasions,
we
have
some
efficiency
requirement,
so
we
want
that
approver
and
a
verifier
doing
education.
They
are
efficient
a
way
to
asymptotically
measure.
Efficiency
has
to
say
that
they
have
to
be
poly
log,
so
any
power
of
a
logarithm
enum.
It's
the
efficiency
in
the
sense
of
how
many
incidents
of
running
time
just
running
time
and
in
this
phase,
since
we
were
communication,
be
small.
A
We
actually
need
to
allow
the
prover
in
this
phase
to
be
linked,
at
least
in
Europe
in
general,
can
be
poly
poly
an
so
a
bit
more
in
a
simple
world
in
this
transition
phase
that
the
prover
is
not
too
efficient.
Okay,
it's
he
has
to
be
slower
somehow
and
if
to
formalize
the
security
proof,
we
speak
about
these
anti
soundness.
A
A
You
would
like
to
have
an
equal
to
an,
but
it's
basically
not
possible
in
products.
So
we
call
this
difference
between
what
the
what
the
prover
is
really
storing
and
Prime,
and
then
the
hey
Matt
one
space
gap
and
clearly
this
is
one
of
the
measure
of
how
good
your
proof
of
space
is.
In
principle,
you
want
space
gap
very,
very
small,
but
in
all
the
construction
that
we
have,
we
see
that
this
is
as
a
cost.
The
smaller
the
absalons
amount,
the
efficiency.
What
for
the
prover
and
the
verifier
depends
it's
it's.
A
It
goes,
it's
less
efficient.
They
are
less
efficient,
so
a
good
if
you
want
to
go
home
and
design
a
good
proof
of
space.
The
first
thing
that
you
have
really
to
have
clear
in
mind
is
that
you
have
to
make
a
proof
of
space
that,
when
this
goes
down
goes
to
zero,
you
don't
pay
too
much
so
the
dependency,
the
efficiency
dependency
own
app.
So
in
principle,
log
or
any
function
that
doesn't
grow
too
much
he
has
to
go
is
impossible.
A
So
not
a
good
in
project
parameters.
The
in
2016
at
TCC
there
was
a.
There
is
a
result
from
random
Devadas.
They
actually
improve
it.
A
lot
on
on
this
factor,
because
we
can
see
that
with
that
proof
we
have
a
spacecraft
of
practical
and
talk
about
practical
parameters.
We
are
of
one
us
and
the
first
one
half
of
the
storage,
so
I
passed
the
proof
if
I'm
say
and
I'm
saving
out
of
the
storage
still
good
the
world,
especially
in
the
cost
and
in
the
context
of
using
real
data.
A
The
goal
will
be
something
very,
very
small,
so
like
1%,
something
like
that-
and
there
is
a
result
from
creep
from
Europe
to
last
year
from
from
ban
that
actually
say
that
you
can
achieve
that.
For
almost
predict
a
parameter,
you
can
go
as
long
as
you
want.
What
is
what
is
called
the
tight
proof
of
space,
and
the
other
thing
I
want
to
say
again
is
that
remember
that,
unfortunately,
in
an
official
space,
we
need
to
require
that
the
prover
doing
the
initialization
phase
is
at
least
linear.
A
There,
an
intimacy
is
at
least
at
least
linear
in
Anna,
so
it
can't
be
super
efficient,
okay
and
other,
otherwise
that
we
have
this
rational
attack.
That
means
that
the
prover
in
the
execution
phase,
if
he
is
very,
very
very
wise,
if,
as
a
very,
very
efficient
installation
phase
what
he
can
do
directly
secures,
we
can
simulate
initialization
again
and
fake
space.
So
this
is
somehow
an
intrinsic
limitation
of
this
construction.
A
Now
I
want
to
discuss
a
bit
more
about
this
Chuco
strain.
So
here
adversary.
We
say,
if
you
think
moment
I'm.
This
is
a
basically
a
trade-off
between
space
and
computation,
because
I'm
bonding
my
adversary
on
both
sides,
but
the
storage
and
the
position
is
bounded.
So,
in
order
to
avoid
to
go
back
of
proof
of
work
and
having
a
ruler
that
you
know
can
save
a
lot
of
storage
because
it
was
a
lot
of
computation
this
is
we
really
want
to
avoid
that?
A
A
How?
How
can
you
do
that
in
products?
So,
if
you
have
a
clock,
if
you
have
application
a
real
world
application
with
a
clock,
what
you
can
do,
you
can
just
put
a
time
a
time
bound
on
the
answer
of
the
prover
during
the
execution
phase.
Okay,
this
is
what
we
call
the
time
model
and
be
careful
that
if
you
want
to
be
in
this
world,
you
need
to
look
at
the
parallel
complexity
of
your
prover.
So
it's
not
enough
to
say
that
he
has
to
do
this
many
step
and
a
step.
A
Think
when
I
say
step,
you
need
to
think
about
some
elementary
operation.
It
can
be
a
call
to
air
and
the
morrocoy
can
be
either
elementary
operation,
which
we
measure
the
number
of
the
complexity
of
the
program.
So
in
this
world,
in
a
time,
module
is
super
important
to
look
at
the
parallel
complexity,
because
I
I
it's
natural
to
assume
that
everything
that
I
am
doing
in
parallel.
They
take
the
same
times.
So
if
I
have
done
operation,
what
I
can
do
them
in
parallel
is
equivalent
to
actually
one
operation
in
terms
of
time.
A
The
other
two
models
that
we
have
here
is
that
I
could
just
say
you
know,
look
at
computation
without
time.
Let's
say
I
don't
have
in
my
application.
There
is
no
way
to
time.
I,
don't
have
a
clock
to
time.
The
answer
of
the
of
the
approval,
so
I
can
use
some
limitation
on
the
computation
can
be
something
that
is
just
infeasible.
So
this
is
a
number
of
running
stats.
That
is
just
we
believe,
infeasible
for
computational
bounded
machines,
or
you
can
have
some
physical
hardware
cost
reign
on
the
prover
I.
A
So
this
is
a
bit
in
someone
somehow
in
a
stream
of
the
other.
They
are
a
bit
difficult
to
use
in
the
real
world.
This
is
the
one
that
we
actually
use
if
I
can,
and
so
you
you
will
learn
more
about
this
later
and
then
there
is
a
third
option
that
we
try
to
use.
We
can't
use
right
now,
so
you
will
see
why
the
problems
Nicola
is
going
to
give
more
detail,
but
is
what
we
call
the
rational
the
cost
model
with
the
rational
prover.
A
A
Before
going
to
now,
finance
professor
of
replication,
just
like
there
are
many
related,
the
definition
about
to
for
spare
is
preferred
stock
solidarity.
Mobility,
I,
don't
want
to
go.
You
know,
definition
of
all.
Otherwise
we
stay
here
day,
but
just
two
quick
things,
I
hope
apt
to
clarify
a
proof
of
space
is
not
the
memory
or
the
function.
It's
not
because
an
odd
function
is
something
that
requires
storage,
just
doing
computation
and
a
proof
of
space,
especially
especially
here
we
are
talking
about
persistent
space.
A
Remove
the
constraint
on
the
prover
to
be
linear,
3,
sorry
to
be
efficient
in
the
execution
phase,
so
I,
don't
field
approver
is
efficient
on
the
execution
phase.
So,
if
you
remember
the
drawing
before
this
is
mean
basically
mean
that
my
execution
phase
can
be
in
principle
as
long
something
like
that
or
like
initialization.
A
This
means
that
clear
is
not
a
proof
that
you
want
to
that.
You
want
to
do
very
often
because
it's
slower
than
a
proof
of
space,
so
you
don't
want
to
do
this
very
often,
but
but
but
removing
that
allow
you
to
somehow
having
simpler
coastal
action
is
a
specific
one.
I
didn't
put.
The
references
already
asked
me
that
I
can
point
it
here.
This
is
just
to
I
hope
clarify
a
bit,
and
now
we
want
to
go
to
use
user
for
space
right
because
that's
the
the
goal.
A
The
first
idea
that
you
have
in
mind
is
that
somehow
you
want
to
they
verify
being
able
to
consistently
check
that
you
are
storing
the
data,
so
it
can
be
that
you
verify
just
send
the
data
to
the
prover
and
he
keeps
it
keeps
a
small
toggle
like
you
can
think
about
a
commitment
to
the
data
and
then
I
randomly
check
that
you
are
still
storing
D
consistently
with
the
commitment
to
a
target.
I
have
their
business
works
to
prove
that
you
are
storing
it.
A
But
we
have
a
problem
here
that
this
data
that
can
be
compressible
and
it's
a
problem
in
general,
because
we
don't
have
any
space
hardness
here.
So
if
I
want
to
use
this
in
the
place
of
a
proof
for
work,
what's
the
resource
that
I'm
using
is
not
clear,
it's
even
more
complicated.
This
is
even
more
a
problem
in
the
case
of
file
coin,
when
the
proven
the
verify
can
be
the
same
person,
because
the
verifier
is
just
some
one
in
the
chain
and
the
provost
same.
A
So
you
don't
have
a
way
to
distinguish
between
the
entities
in
the
blockchain
and
then,
of
course,
for
sure
the
data
can
be
compressible,
so
there
is
no
space
at
Arden
s
at
all.
So
how
can
you
solve
this?
This
means
that
basically,
okay,
proof
retrieve
ability.
This
is
important,
but
we
can't
give
up
this
proof
of
space.
So
you
want
to
compose
the
notion
of
before
a
proof
of
space
with
the
proof
of
receive
ability.
How
do
you
can
compose
it
again?
The
first
idea
it
could
be
just
two
both
of
them.
A
Okay,
here
the
prove
the
verifiers
and
the
data.
The
proof
runs
both
the
insulation
phase
of
a
proof
of
space
and
he
has
the
data
and
and
then
I
check,
both
the
data
and
in
a
natural
way
and
I
run
the
execution
phase
on
my
proof
of
space.
Is
this
okay
nope
for
the
same
reason
as
before,
because
I
can
still
compress
these.
A
So
in
some
reason,
okay,
you
are
proving
space,
but
you
are
proving
space
in
a
very
inefficient
way,
because
if
I
can
compress
these,
this
is
the
only
things
that
give
me
space
hardness.
So
honest
players
are
going
to
use
two
times
the
space
but
bad
players.
Dishonest
players
are
going
to
compress
these
so
I,
using
these
so
I've,
almost
1/2,
again
I'm
back
to
one
out
of
space
gap.
A
That's
not
a
good
situation,
and
it's
even
worse,
if
you
put
these
in
the
contest
of
a
storage
provider
and
I,
want
to
somehow
use
these
also
to
guarantee
some
replication.
Some
redundancy
redundancy
property
amazing
that
the
verify
is
asking
you
is
asking
to
the
prover
to
prove
to
store
two
copies.
So
this
is
known
is
a
cheating
prove.
I
can
always
pretend
to
store
two
copies,
because
it
does
two
execution
of
the
proof
space,
but
just
one
of
the
proof
of
retriever
ability.
A
So
all
these
spaces
save
it
and
the
more
the
worse
here,
okay.
So
what
this
teaches
us
is
that
a
good
solution
needs
to
be
a
protocol
T
simultaneously
on
the
same
space
on
the
same
data,
a
proof
of
replication.
Oh
sorry,
approval
retrievable
in
an
approval
space,
okay.
So
this
is
how
the
proof,
the
definition
of
proof
of
retrieval
replication.
A
Born
okay,
so
if
you
want
to
be
a
bit
more
technical,
basically
now
the
verifier
send
this
data
I
have
only
one
operation.
At
his
run,
we
call
it
the
replication
I
replicated
Eddie,
so
I
get
this
R.
That
is
land
and
if
the
place
some
of
us,
it's
my
fear
of
the
disk,
but
it
contains
the
data
as
well
and
the
verifiers
just
attack.
A
This
is,
can
be
an
interactive
or
non
interactive
process
and
then
I
have
the
same
idea
of
there's
a
Christian
before
so
I
have
an
efficient
phase
where
the
approval,
sorry
in
a
verifier
can
rapidly
query
the
the
prover
to
chakras
is
still
storing
a
large
fraction
of
our
and
our
contains
the
data,
and
we
have
a
third
algorithm.
If
you
compare
these
two
prefer
space.
That
is
very
important.
A
That
is
a
struction
I
need,
since
my
data
are
inside
are
now
I
need
to
be
able
to
go
back
and
extract
the
data
if
it's
needed
and
a
clearly,
we
want
this
efficient
as
well.
So
a
practical
efficiency
of
this
of
this
part
is
actually
one
of
the
main,
interesting
open
problem.
Research
question
that
that
we
have
so
if
we
want
to
formalize
the
definition
they
the
idea,
the
goal
is
that
you
want
the
past
soundness
plus
plus
some
a
couple:
a
K
replication
security.
A
That
means
that
if
my
data
are
divided,
let's
assume
that
the
data
now
has
two
blocks:
D
1
and
T
2
I
want
them
be
replicas
as
the
same
structure
such
that
from
each
block.
I
can
reconstruct
the
data
block,
because
this
gives
you
immediately
these
the
replication
property
in
the
storage
when
you
use
it
as
a
in
the
storage
provider
setting.
A
Unfortunately,
this
is
is
not
possible
to
achieve
in
in
a
pure
sense,
because
the
dishonest
prover
can
always
support
us.
The
replica.
This
is
what,
in
the
poly
paper
in
that
interest.
These
primitives
is
called
a
little
sabotage
attack,
because
if
even
if
I
have
this
structure,
what
we
can
do,
we
can
manipulate
the
replicas
and
store
a
2
plus
a
1
in
the
place
of
virtue
and
I.
In
this
case,
what
is
really
happening
is
that
ok,
this
is
the
replication.
I
have
two
things,
but
I
can't
extract
anymore.
A
My
second
block
in
the
data
just
in
looking
at
this
I
have
to
look
at
this
stain
another
way
around
if
I
lose,
this
I
lose
both
also
the
second
piece
of
my
data.
Ok,
so
I.
Somehow
this
redundancy
replication
is
destroyed,
ok
and
there
is
no
way
to
really
really.
There
is
no
way
to
prevent
this.
A
dishonest
prove
I
can
always
do
that,
because
it's
doing
this
on
his
own
that
all
right
and
it's
very
hard
to
detect
something
like
that
with
the
checker.
A
So
the
idea
here
is
that
to
use
some
rationale
proven
go
again
in
the
rational
setting
and
say
you
know,
even
an
adversary
that
does
this
so
does
this
kind
of
sabotage
attack
is
not
going
to
save
more
space,
that
an
attacker
does
everything,
but
not
the
sabotage
attack,
so
I'm
saying
that
sabotage
doesn't
help
you
to
save
space,
and
he
actually
put
you
with
some
risk
because,
as
I
told
you
before
now,
if
you
lose
these
sorry,
if
you
lose
these
you,
you
lose
everything
right.
So
it's
not
rational!
A
It
doesn't
allow
you,
so
you
are
not
going
to
do
that
in
principle.
Proving
this
kind
of
property
is
a
bit
tricky
because
you
you
in
a
formal
way,
you
have
to
show
that
for
any
prove
or
that
does
this
sabotage
attack,
there
is
another
prove
it
then
for
any
adversity
does
the
sabotage
attack
marries
another
doesn't
do
it
and
it's
equivalent
it's.
It
can
be
painful,
but
thank
you
man.
He
did
all
the
hard
work
for
us
and
there
is
a
shortcut
to
this
that
you
come
to.
Take
your
proof
of
space.
A
Just
prove
that
is
that
professor
Hughes
proof
of
space,
and
also
that
is
a
property
of
ability
that
is
usually
easier
to
prove
then
than
the
former
on
a
committed
value,
and
you
are-
and
you
are
done
so
some
something
about
if
you
want
to
read
more.
This
is
this
is
like
they
say,
the
technical
tools
that
we
have
okay.
So
it's
super
impressive.
For
this
reason
it's
super
important.
A
A
This
is
some
reference,
as
the
one
said
before
there
are
some
technical,
wrappers
and
paper
work
in
progress
paper
of
protocol
lab
and,
and
there
are
the
two
papers
from
from
ban
they're
both
available
on
any
printer.
So
if
you
want
to
read
more
about
this
stuff,
it's
yeah
now
we
can
do
two
things
I'm
asking
to
the
other
speakers
and
to
everyone.