We're going to help.

Me run home.

But we're going home already.

Mom.

And also.

For.

The long haul and today's.

Security concerns for the older one.

Are now thanks very much.

Happy to be here and invited when I
was talking to Maria she said Your

title is very cryptic so it's not crypto
it's just cryptic so I'll try and

hopefully make it a little bit what I what
I had in mind and what's behind the title.

So something about me I started network
operations in one thousand nine hundred

four then moved to engineering in
architecture worth different rules for

something like twenty years always
pretty close to operations in one way or

another and
security isn't my main focus and

also hopefully this is a bit of a meeting
of different cultures and communities.

In this.

Community audience so.

None the less I'm pretty involved
in security one way or another.

To give a little perspective.

Back when I started.

There were a whole bunch of things
that were different that is now so

we had thing called finger which was so
they ran on a lot of UNIX machines

basically it was plain text transport you
could see who was logged into a machine.

You know privacy
considerations all over the.

Place but that was the state of the world
back then you know it was a small

community S.S.H. didn't exist
there were ways to do encrypted

sort of command line sessions but
it was pretty rare

at near network I was working at the time
working for B.B. and B.B. and knock.

There was open privileged you know sort
of top level telnet access to routers

it was open because the idea was
it's a community who all wants to

to understand you know what's what's
working what's not working so we made that

available that's definitely not the type
of thing that would happen these days.

We would use sometimes in troubleshooting.

Reachability issues we would we would
use something called the The percent.

Percent hack where you could say if you
wanted to instead of saying user at

domain you could say user at domain

a user percent domain at some other domain
see if you couldn't reach their S.M.T.P.

server you could bounce it through
an intermediate one that you had

a sense might be able to reach it and
that was you know that was fine and

very helpful and useful definitely not
the type of thing that happens today.

You know sort of talk about
the updates to some of these things.

Further down in my talk
in terms of questions

I think maybe better to save them till the
end which I hope there will be some and

have a discussion but that's probably
the easiest way to get through.

There was also for

trace there was trace route with a loose
source option where you could say.

Tracing to a destination but
use this intermediate hop to get there.

And that was important for us because
that way we could if there was it

reachability between my network when I
was trying to reach another person's i.

Could bounce it through a third network
that I thought might have reachability so

I could see how they were connected and
sometimes it was a way to also

troubleshoot you know what what those
people the people on that network may be.

Happening might be seeing and so

the reason that it's basically doesn't
exist today is that it's kind of

deed possible de DOS vector because
those loose source packets are what

they call process which they're switch in
software versus switched in the hardware

in the in the fast path of the router so
that definitely doesn't exist either.

So obviously we've seen a lot of
improvements since then not all have seen

employment Why is that and that's kind of
really the upshot of the talk today and

that's why I called you no way why
I gave it this cryptic title so

I'll talk about a bit about
the multi-stakeholder worlds and

then some examples of how it applies
to security concerns with routing with

spoofing of different types with D.N.A.'s
Domain Name Service and hopefully.

You know I am sure people have a variety
of different backgrounds here and

hopefully I'll be bringing
people to some so

Munder staying home I mean
I'm sure it'll be sort of.

Well understood for a lot of people
maybe some others not so much but.

In the multi-stakeholder system

you know what I had in mind is
there's a variety of actors there's.

Development protocols by the standards
organizations like I.T.S. Tripoli for

net and
lower layer network things W three C.

obviously for
web protocols hardware suppliers

you Cisco's your Juniper's Apple Intel
you know Intel is obviously not a retail.

Organization but

they're inside a lot of things as
there are their ad campaign says.

And then there's software also
router software there's a Wes's

all kinds of things and for

operators there's interest service
providers there's content distribution

networks enterprises including you
know universities governments.

Then there's obviously the users
regular people the military

your grandparents my grandparents or
my my in-laws.

And then the government you know
has this has a piece of it there's

legal issues regulatory law enforcement.

The military again.

For education it's universities
vocational schools secondary and

you know primary schools to
you know educate you all or

or you know the citizenry and
the workforce to to get get

us where we need to be nongovernmental
organizations like Internet registries

advocacy groups of different
types they all play a part in it

as you may expect they've got
various incentives for what they.

You know what they do and why.

So.

You know some dimensions obviously
time always takes too long money

it's too expensive effort it's too
difficult complexity makes things

harder to understand and manage you know
maybe makes them more prone to failure.

Risk never sure what you know is
what's the outcome going to be.

And.

Then entropy hard hard to keep it clean
once you deploy something how do you how

do you keep managing it through its life
cycle without it just sort of decaying and

leading to more mass.

So we think about.

Some things that you know let us hear
also the robustness principle as as John

pastel articulated it be conservative
in what you do liberal and

what you accept from others so this was
something that he wrote in a in an early.

Specification for T.C.P. IP protocol but

was you know kind of
the governing motto or idea for

a lot of the development of the early
Internet even through today you'll get

many good things obviously it makes
bootstrapping and interoperation easier

keep the internet running and growing
you know the easier it is to deploy and

start up things you know it's growing.

Obviously might have yielded some of
our current challenges was a different

time when the community was smaller
everyone knew each other better.

You know the early Arpanet and
Internet days so.

Obviously now we're we're we're
more cautious more guarded.

We you know want to to.

You know not necessarily trust
by default that's part of it.

So.

Let's talk about something simple and

like I said I prepared this I wasn't
sure exactly what the audience you know

what the background and
experience of it would be so

hopefully it's not too
remedial you know helps bring

everyone into it since you folks are more
security minded than network or is it.

The fundamental building block block of
internet routing is on antonymous system

so it's a network under a single
administrative control.

Uses I.G.P. on the inside E.G. P.

to the outside so you know if you make
a diagram it's like a cloud right so

inside it's among the different routers
uses the internal gateway protocols.

They handle top ology discovery and state.

State management so figure out where
your links are between different nodes

what their capacity is and
whether they're available or not available

you would deploy it inside a P.C.
an enterprise university etc

There's you know several different
examples that there are use B.F.

and I.S.A.'s are the biggest
ones rip not so big anymore.

Then on the facing the outside of
the autonomous system between autonomous

systems we use external gateway
protocols you know generically speaking.

This is how the different.

Routing domains advertise
reachability between each other and

so is really the only one currently
in use there was an earlier

when I started that was you know I
knew of one place that used it but

it was basically already on its way out.

This is right how they have
the network speak to one another and

the fundamental relationships they
have between each other we call.

Customer or you know all the transit or
non transit in terms of does the traffic

traverse they to go to another place or
does it does it terminate in that

so sometimes people call that
a customer relationship so

it's called settlement free
interconnection or appearing.

You know the customer thing
focuses on the economic

arrangement but really from a technical
point of view it's really does the traffic

traverse that domain to go other places or
does it does it drop their.

Forwarding decisions are based
on destination IP address so

advertise rather advertisers
reachability for destination for

a network traffic for that doesn't ation
then flows toward the advertising router.

And source IP isn't checked by
default I mean in general so

that's the kind of big picture of it it
will talk a bit more about that later too

so what are the kind of risks that
we've seen control plane attacks where

you're attacking the routing information
flow control plane is what we

think of as the you know where
the routing information gets exchanged

in the data plane is the you know is
where the data actually traverses so.

There's.

This origination leaks also
are probably the bigger concern and

those can result in interruption
of traffic flow more into detail.

Coming up.

So what are the risks that we see
within the control plane they're

bringing down B.G.P. sessions.

Either spoofing the source of IP source IP
and that sending a T.C.P. reset because of

functions over T.C.P. connections so
if you could set send a T.C.P.

reset which basically says to your to
the other end of the connection close

done you know if you could do that then
you could just bring it down you know so

they'd be exchanging traffic as they bring
it down and all those routes would drop

the effect would be service
a bunch of solutions were.

Developed for this in the.

B G P T C P M D five which
basically involves a shared secret.

So both sides enter the same password and
then they.

Encrypt the data between them so

that somebody without knowledge of
that shared secret if they sent a.

They sent a message including a reset
it would you know when affected the.

Router and.

The you know one critique of this
approach was OK what if somebody

just sort of sends a bunch of of
known bad five known bad packets

causing the router to have to run the M D
five calculation on a bunch of stuff that

it doesn't end up caring about you know
they could basically do that all day long.

It could overwhelm the C's C.P.U. and
you know deploying the security solution

could actually make things
more vulnerable was a concern.

Then there is another solution
developed called generalized T.T.L.

security mechanism which.

Normally be the packets because
they just run between routers that

are right next to each other by default
they have a an IP time to live of one so

that basically it never
has to go further than.

Than one hop away and

you know the way traffic goes
every pop Decorah meant to T.T.L.

then when the T.T.L. is reaches
well when it arrives with a T.T.L.

of one so you did create Did you
know Decker meant by one zero so

it's never going to get forwarded so
the idea is if somebody sent

a packet from further away which had
to have a larger T.T.L. to reach

the destination it would arrive with and
you know a T.T.L. other than one and.

This would be a protection by
making the router enforce that

that property it would
protect you from somebody

out in the Internet making
an attack on your on your router.

Basically you know that's seen
a little deployment not huge so

why are these security mechanisms that
came up the were developed haven't

you know they haven't been universally
deployed by any means still attacks

haven't been seen before you know or
after because and and wide.

It's not totally well understood
it's hard to guess the T.C.P.

sequence number combination
of ports you know because.

You know there's there's
a large numbers both sides.

You know the initialization
sequence number initialization

is pretty well done now so
that's one theory of why the attacks

haven't been widely seen even though
many people don't deploy these

protections also maybe because there
is Don't want to you know just bring

down the Internet because that's
what they used to to do what they do

the other large class or even bigger
class of of problems with P.G.P.

is Route leaks or mis origination So
basically what happens

with the risk risk is that you'd
be black holding some traffic or

that you could you know divert it
through your network and inspect it.

Or you could overload intermediate link
by sending more traffic there than than.

It's designed to handle.

Generally what we see more often
than not is unintentional.

Problems where the first one
that I remember which may be

the first one that's generally well known
or thought of is in one thousand nine

hundred seven there was the A.F. seventy
seven thousand and seven incident so

what happened was this Thomas system.

Just generated tens of thousands
of routes which was you know about

what the internet routing table was back
then it was probably about at the time

forty to fifty thousand routes now it's
well over six or seven hundred thousand.

So basically it's thought
that this was due to a bug or

somebody you know enabling a sort of
hidden feature that they didn't realize.

And what happened in my
network at the time and

generally is that all the routers
ran out of memory so.

You know it was a very day for
the Internet and the only way to clear

the the problem was to reboot the routers
either you had to send somebody outside

the facility to look at the power or if
you had done access through a phone line

to a modem to the console port of
the router you could tell us a reboot.

So that was a very bad day and
some variation of of

route explosions you know have come up now
there's sort of more safeguards usually

people configure maximum number of
prefixes at a session level that you're

expecting to see so if somebody goes
over that it tears down the session so

you can't you know this so
so as as far as getting

a network wide Internet wide problem
that doesn't generally happen.

There's also what happens more often is
misconfiguration because there's poor.

Really you know the software
by default will allow

you to configure session and for it to
come up without any filtering in place so.

That's you know a lot of times
people will not configure

the whole piece that says no maximum
number of routes or only allow these

fifty you know twenty or fifty or one
hundred routes to come over the session.

But you know just it'll come up and
a whole bunch of routes will come in that

aren't expected so that's
the unintentional So a lot of times.

There's Mr origination which is
a different thing where you you.

Accidently originate a route in B.G.P.
that doesn't belong to you.

And obviously then attract the traffic for
that that network.

Depending I mean sometimes
you'll only attractive for

your neighbors sometimes you'll attract
it for a broader part of the Internet

usually it's accidental it's just
a typo in some configuration.

Some incidents that have happened there
was an incident with Pakistan television

where they had an internal censorship
system that's what surmised

you know no one who's authoritative
as you know told the whole tale but.

So they would advertise routes for
you too because they didn't want.

Either their internal people this to
be able to use You Tube or they wanted

to be able to filter certain content and
so at least passed the borders to.

Peas and you know a lot of Youtube
traffic a lot of Youtube for

a lot of people was unreachable that day
or those couple hours or whatever was.

There was a China incident in China where
they had they leaked some addresses for

some of the root name servers
it's thought that that

was probably unintentional they
may have had some internal sent.

Your ship set up or or
what is not necessarily known but.

For a little while traffic for those
routine servers was going into China.

Didn't last very long so intentional
we've seen it you know a few intentional

problems or apparently intentional
problems there was this so-called Latvian

hijack I mean no one knows for sure if
it's was a hijack but Renaissance was

a company that did the analysis where
it seems that you know there were.

Different episodes over the course
of weeks or months that you know for

a little bit of time where some traffic
was diverted to Lafayette and then.

You know sent on its way to its ultimate
destination so it was in a totally

disruptive mis origination thing so
still unknown there was a bit coin attack.

Where the seemingly
clear motive was to gag

some parts of the Bitcoin
infrastructure temporarily so

that you could gain an advantage to
make more money or whatever over that.

Might have worked then there's other
things where people use a little

bit of on a new used address space they
advertise it temporarily and then they.

Send a bunch of spam or
they you know do some other

you know bad thing because that
address space is you know it's not

theirs it's not attributable to them you
know they can use the good reputation or

the neutral reputation of the of
the owner and then go away.

And so that that happens from time
to time less record of it you know

usually the scope of that of how
far that those originations go

is in the global routing system is kind of
variable so it's not always observable.

So how do you know what's
the way to combat this is

is how to know when
the routing is correct so

how do you figure out when the autonomous
system belongs to us you know a particular

person when you're on the phone with them
setting it up or exchanging e-mails.

You know how do you actually even know
that the person on the other end is

representing who they say they are that's
probably not a huge problem but but

you know points up one kind of thing that

you know part of the part of
the overall ecosystem so.

Then you have an idea of
OK what address space.

What address space is an autonomous system
authorized to announce So you you know

generally white list the routes you allow
them in through a filter configuration.

So you explicitly lead in those routes
which implicitly denies the other so

that's you know a protection that people
generally use but you know would have

prevented some of the earlier problems and
you know so it's not globally.

Globally used and then with a S.

path filtering you can white list
the transit relationship so you can say

all right we expect these autonomy systems
to be on the other side of this one that

we're connecting to but you know there's
a limit to what might be over there so

that you don't have a situation where
they're advertising traffic for a.

For a for a whole network that they
that they are not authorized to.

The challenge is we want
the system to be dynamic and

flexible we want to you know have
minimal maintenance not have to touch

every time somebody wants
to you know bring up a new.

You know somebody who's not directly
connected to you wants to do something

you don't want to have to you know
contact everyone else in the Internet and

say OK I'm making this change so
you know that's kind of the challenge.

But you wanted to be too flexible right.

So what are the different you know back
to the multistakeholder piece What

are the different in addition
to obviously you know I talked

about the protocol developers and some of
the security solutions the hardware and

software makers that have to implement
them and then you have to acquire and

deploy that then you've got
these other characters like

regional Internet registries which
are fundamentally important because how

do you know you know that
people don't reason the same.

Number resources IP on a system number
resources through a registry and

so there's five of them in the world
roughly Continental Africa

lacnic is Latin America which is
basically everything south of of the U.S.

in the Americas right which
is covers Europe A.P.

Nick Asia Pacific region and Aaron
which is North America and Caribbean.

Their membership organizations

they issue these autonomous system
numbers the IP address numbers for

V four of the six called you know
we call them blocks or prefixes.

Their members and
the people who they issue

resources to are you know they have
a relationship with the you know try and

figure out who those people are and
you know how how to.

You know that they're legitimate so.

You know we call it a business
relationship that may or

may not involve money that's sort of
outside the scope but you know there is

a relationship between the organization
and the people who use it.

They have to keep their
contact up to date because.

They have to know how to
get in touch with you and

you know if there's an abuse situation or.

That kind of thing and

if somebody else wants to report abuse
they have to know you know how to how to

get in touch you probably heard of the WHO
IS service which is you know a mapping of.

Resources to organizations then there
is the internet routing registry so.

There's this is basically
a system that will you know and

here's some examples some of them are run
by the routing registry by the Internet

registries some aren't and
they express routing policy or

intent so the mapping of of a prefix
to autonomous system and they can even.

You know with the language
it's called R.P.S. All right

policies specification language you can
even describe neighbor relationships.

They I would say it's it's.

You know one uniformly used you know so
not everyone uses it.

Exactly how you're supposed to
use it isn't you know uniformly

done there's no expiries So
like if you put something in.

You know if you do something for
a particular customer or

you know somebody submits a.

Piece of information to the Internet Rajan
registry it stays around there whether or

not it's still in effect you know it
just can stay around there forever

there's no particular motivation
to clean it up and so

people have a lot of issues with trusting
the the information in there and

feeling that it's fresh and
reliable Also there's no linkage of

what you put in there with your right
to speak on behalf of that resource.

You know broadly speaking right for
instance and I think A.P.

Nick in their routing registries they
say OK if you're going to put something

in writing registry about a particular
prefix you have to be the person who it's

to whom it's issued so
they have at least kind of

the authorisation piece baked in but
but that's not uniformly true.

So Dunder stand a bit about
the challenge of this whole thing

think about you know the what goes through
when they provision B.G.P. customer so

the customer details come from the sales
order or some sort of documentation and

you figure out OK It says it's going to
use this autonomous system number and

what prefixes are are you know
are we expecting from them so

there's no way to know that from inside
the G.P. itself you have to go to one

of these other pieces you
know places these registries.

But sometimes the documentation
is big US because it'll say

you know there's a mergers and
acquisitions and name changes and

stuff so if it's it if you're
dealing with a company called A.B.C.

Inc And then you go to registration
it says already been choose dink

chewed incorporated into a human
can say OK that seems like

the acronym matches you know
seems like the same thing but you

know from a machine readable computational
perspective it's not it's not that.

Reliable.

And then you you know.

How do you understand what the downstream
A S is are that's that's hard to do too

and then if somebody is bringing
a piece of address space that they say

yes Somebody lent it to me
we have an arrangement so

you can override this registration
with what we call letter of agency.

There's no cent standard for it like you
know it's a letter on some letterhead

which I mean people who even use that
many more so then you get a P.D.F.

Well those can be forged and
that's I've seen that happen.

So if you if you bring
up if you're working and

I esteem you're bringing
up a session two and a.

Non-customer.

So it's a non transit network you.

Generally don't enumerate
specifically what

prefixes you're going to hear because
that can change you know it's them and

their customers so you don't necessarily
want to hard code it on the router because

it gets difficult to manage and you know
sort of inhibits the flexibility and.

Dynamism of the system so
generally what's done is just try and

make sure that some large that they
don't send you routes going to some or

other that large non transit network so
that you know and

how do you know which these are is
just through experience so if you say.

You know I work for Comcast.

I'm going to set up
a connection to eighteen T.

Well I think that a.

I'm pretty sure they should never
advertise me Verizon's route so

you know I'll filter
anything that seems to be

you know that seems to say that
Verizon is downstream of a T.N.T.

from that connection because I don't think
that should ever happen but there's no.

Ground Truth for this is no way to
verify that you know what the proper

relationships are supposed to be and
then the changes

you know a customer will call up
they'll say change the set up because.

You know we're making some changes
you know we can handle that but

you know it's it's.

A bit a bit onerous there's ways
to automate it using if they

if they average if they maintain their
own stuff in the routing registry

we can you know query that it intervals
and really generate our filters and

reapply them so
there's ways to automate it.

Workflow still all the thing.

As I mentioned before doing those checks.

Mergers and Acquisitions slows
the dynamism So what's a.

You know from the standard side
what's a proposed fix to this so

our peak is called Resource
public key infrastructure so

it follows the the resource
delegation hierarchy so

you get when you get a resource from.

Regional Internet registry they you
can basically through their system.

Put together you know register binding
of origin A S to that prefix and.

They use X.

five zero nine certificates so you know
it's sort of cryptographically verifiable

that the person who who who put it in is
the you know correctly authorized party.

But it can also work in
a delegated system where

they can say you can set up your
own certificate authority and and

your own but very few people do that
the system generally hasn't seen a lot of.

Of.

Uptake.

So also just to finish that
you know the validation or

sort of the route origin verification is
basically you know you have a another

piece of software that pulls in all these
are difficult runs all the operations to

make sure they're you know seem correct
correctly signed and delegated.

There's you know this is another case
where the people who make the software

another stakeholder well who's
motivated to make the software there's

three versions of open source versions
that exist there's no commercial support

you know being a company a large company
you know might take that take that

on smaller companies there they're going
to be like you know we're not experts and

that if it breaks how do we get
it fixed questions whatever.

Tools to diagnose the problem are yet
develops.

Certificates going to expire so
if if the if you signed a bunch of these

these are called bro is
Route origin at stations but

basically if you signed a bunch of these
cryptographic objects they always you know

it's they have some experience could
be a year it could be twenty years but

you have to keep that in mind and
you know remember to refresh or

what have you reassure
the certificates when that time comes.

Scaling properties are known.

And what if a government says you know
goes to the the right registry and

says we have a takedown notice for this.

For this network.

Issue a you know remove their origin.

At the station so that basically anyone
who's running that software will no longer

accept traffic for it you know people
are concerned about that because that's.

Governments have ways to do different
things on the Internet now.

But this is a whole new level of
breaking the contents of any that

that they they have been using before so
that's of concern right.

Expertise varies wildly widely I
should say you know enterprises

don't have the expertise speeds do so the
enterprise are usually going to say well

what you know I'm paying you a law for
this I don't want to know about all that

complicated stuff you guys figure it out
and fix it and you know do it right so

you know there's there's a stakeholder
Who's your customer who is telling you

you know they're interested to a point but
not too much.

All right onto a different
class of risk spoofing this is

injecting unauthorized traffic you
guys probably have some awareness of

you know source IP address spoofing e-mail
address spoofing there's you know these

are well studied abuse vectors
in the in the security world.

You can inject bogus data
you know guessing T.C.P.

sequence numbers you can you know insert
bogus traffic into existing streams

there's fixes to it like improving
the secrets number generation software

vendors have to code the fix you does have
to upgrade the software more stakeholders.

The risks are implication you know here
are some examples that we've seen in

history this so-called Smurf attack was
one of the earlier ones early in my career

that we saw that where people would send
up a packet with a certain address would

generate like a ton of traffic you know
low investment by the attacker high

yields of bad stuff happening you know
rather code was upgraded people people

deployed so that that's pretty well fixed
S.M.P. amplification attacks you know

another asymmetric thing where somebody
sends a packet receives a large response.

You know that there were home routers that
had this problem this is only going back

a few years and OK so what you know
what should the I speed do we don't

own the home router you know in a lot of
cases so should we block this traffic

because we think this is not something
people would ever want so you know.

They're only so motivated to fix it or
even understand it you know should

a dynasty block that by default to
you know improve the the ecosystem.

It's you know an open question.

So.

Another solution is to block source
spoofed source IP addresses so

this is you know a best what we call best
common practice which is another kind of

R A C which says you know
block at your network edge

you know traffic that has a source IP
address that's not on that subnet and

you can do with an access control list
there's router features that can ease

the administrative version so they
basically reverse the routing they say if

if I would route to that destination
on this port allow the that source

you know something with that source
IP address to come in but if it's.

But if it's not drop it so you can do
that at the edge of the network but

if you're kind of in the core you but

not know where you know where all
the different possible paths routers route

traffic can enter because like I said it's
forwarded on the basis of destination so.

Even though I might route this way to
reach a certain destination they may route

this way to reach me so
the the routing and

forwarding planes don't
necessarily line up.

You know OK so I took talked about this
apology considerations where the stubs

form is considerations some router
hardware couldn't do it at high speeds

that's pretty much not the case
anymore with modern stuff but

you know flexibility
still want to keep that.

Complexity.

Source spoofing people
have all seen spam and

that kind of thing you just
consent to any destination.

What's a way to combat that.

So.

Now we can.

Close what so called open relays so
that people you know are basically using

a really outbound that they're
expected to be using us and S.M.T.P..

And so if you're a customer and
I are an A.

Or a an enterprise you'll have some
password you know credentials that you can

use to submit mail and then that will send
it directly to the destination I mean

it's not so common to see many many
many hops of intermediaries anymore.

You know among different.

Entities handling entities So
basically the open relay thing

is been closed through software knobs and
so forth.

There's also you know user sort of
community developed blacklists and

so forth you may know
about there's the mail.

Abuse working group where an industry
group of industry practitioners came

together and said OK we want to try and
improve this situation solve these

problems how can we come up with
best practices and so forth.

Then there are some standards based
one sender policy framework D.

Kim these are ways to say what

hosts can send on on behalf of
a given domain whether authorized.

But there's legitimate uses for
changing the source.

You know of.

The piece of e-mail like you might want to
send something on behalf of somebody else.

You know if you if you're sending you know
a huge e-mail campaign you might one do it

from your laptop you you
send it to a third party but

you want to know who they want you want
the people receiving it to know who it's

from notionally if they were and
if they were applied for it to go to you.

Guinness Book of things is the last
class of things they'll talk about

so you know probably all know
is the Domain Name Service So

it's how you use how you know how to
map a name to an IP address right so.

That things that attacks that exist and

it's been seen there are cash
poisoning attacks where people inject.

They inject extra data.

Through a.

You know utilizing the some some form
some combination of guessing and

other things to inject incorrect
information into caching servers so

users query caching server for
an answer and that it's possible for

an adversary to to put that information
in there that will end up getting

served to users and take them away from
where they think they are supposed to go.

D.N.A. is the main proposed
solution to this scene

you know variable extremely limited.

Course fairly limited deployment and why.

So you know if it provides
a clear authorization chain for

validation using cryptography and
so forth but.

There's also residual risks

their client cache communication
is still not protected.

Also has potential for.

Amplification attacks because you can
submit a relatively small request and

get like a mountain of stuff in response
and if you spoof the IP address

of who made the query then that
huge response will go to the.

To the victim let's call them and so
it's another case of well is the cure.

More problematic than the disease.

Also sometimes people miss
configure it shooting.

Selves in the foot where they
did miss configured it and

since it's a D.N.A. protected domain
if they screw up the configuration

it'll invalidate everything that's
coming out for that domain and

so you know it'll take them off the off
line and we've seen that happen

with even go you know governmental
agencies who've messed it up so

the way to work around that is to deploy
what's called a negative trust banker and

basically sort of say OK we're going
to we're going to a white list

the things from that domain even though
they seem wrong because we don't want

to like all be unreachable and
broken so it's kind of a workaround.

And those are the main things
that I want to talk about.

Just to wrap up the you know.

The idea was to talk about how many
different parties are involved

in network security you know what
the lifecycle is and you know.

The cryptic title was you know another
way of saying why aren't things fixed yet

so with that I'll take
questions if anybody doesn't

make any sense.

Right.

So that.

It's like.

Right.

Right so the question was about.

If I think I understood right how do
you guarantee that the delivery path of

the traffic is what you intended to be and
not diverted through some third party so

even beyond R.P. kid R P K I R O V.

Rad or Arjun validation
depending on R.P.I. there's.

Another B.G.P. security protocol called
B G P SEC or path SEC path security

which attempts to validate the not just

that the origin is correct on the B.G.P.
route but that the path that.

It's in the the route as it as
a reaches you is correct and intended.

This is it even more complex thing to get
right especially given the number so it's

a it's a standard that's just recently
been approved in the I.D.F. but its.

Its deployment prospects I'd say are very

unclear because it's more
complicated to get correct.

It's interplay with the various ways
people have deployed and used P.G.P. are.

Unknown because it may not match
all use cases that exist so

you know you're right to bring up that
point and people have shown even with

the route origin validation that
there's ways to attack it and

since it's only checking the last
a few If an attacker can can basically

forge a message with the you know correct

terminal A.S. it won't necessarily be.

You know it will be considered valid by
the by the router that receives it even

if some of the intermediate
things are not correct so

it's a residual risk to that
I forgot to mention but yes.

Any other questions.

You know everybody has to get the classes.