0:00

It's time for security now. Steve

0:02

Gibson is here. You'll talk about

0:04

Gps, fussing, how it works, what

0:06

one can do to avoid it.

0:09

You've all heard about that Vpn

0:11

flaw that Ars Technica says makes

0:13

all Vpn useless. Not so fast.

0:15

Steve explains why it is not

0:17

anything to panic about and then

0:19

speaking, I'm not so fast. Google

0:21

has stopped progress on abandoning third

0:24

party cookies. Steve now knows why.

0:26

He will explain all that a

0:28

whole lot more coming up. Next

0:30

Security Nasty. Room

0:33

I guess you love

0:35

from people you trust.

0:39

Was his tweet. This.

0:45

Is Security Now is Steve Gibson

0:47

episode Nine Hundred Seventy Three Recorded

0:49

Tuesday, May seventh. Twenty Twenty Four.

0:52

Not so fast. It's time for

0:54

Security. Now the show. We cover

0:56

the latest security and computer news

0:59

and privacy news. And of course

1:01

it was. I find Tv thrown

1:03

in with this guy right here.

1:06

Steve Gibson, the arbiter of all

1:08

that is good and kind L

1:10

as well. Go for that. Yeah,

1:13

Elo Leo. Although here yard the

1:15

beginning. Of May. And.

1:19

As I promised. I. Did have

1:21

some time. To. Dig into.

1:24

The. Issue that came up

1:26

actually two weeks ago. When.

1:29

In the middle of the show you said

1:31

hey, Google just. Change

1:35

their plans on third party cookies and I

1:37

said. But. Ah, Anyway,

1:39

so. We're going

1:42

to talk about that. The

1:44

today's episode is titled Not

1:46

So Fast. Which. You

1:48

know, As if that expression

1:50

not so fast there. which

1:54

is what the uk as they do google

1:57

but we're going to first look at what

1:59

danger is bruce presented by the world's current

2:02

and growing dependence upon GPS,

2:05

and why is that any concern, has

2:09

the sky fallen on all VPN

2:11

systems as the tech press has

2:13

been reporting since yesterday when a

2:16

blog post really went a little out

2:18

of control. I was really hoping, and

2:20

I wanted you to explain option 102

2:22

or whatever. Option 121. Yes,

2:26

I really want to move on. We'll know all about

2:28

that by the time we're done today. Thank you. Also,

2:31

a couple

2:33

questions more from our listeners, still

2:37

bogged down in what is arguably

2:39

a quagmire of network authentication options.

2:41

So I'm going to spend a

2:44

little more, continuing to come into

2:46

CRISPR focus for me, so I

2:48

figured let's spend a little more

2:50

time on what's going on there. Also

2:53

we may have an answer to what

2:55

Apple was doing with the iCloud

2:57

keychain deleting and what was going

2:59

on, something that absolutely makes sense,

3:02

so we're going to cover that.

3:05

And also, finally, as I

3:07

said, I invested

3:10

no little bit of time. You'll

3:14

hear the term bureaucracy used

3:18

more times probably than any large word

3:20

in this podcast

3:22

because, boy, I guess

3:26

any kingdom that's been around as long

3:28

as the United Kingdom has

3:31

continued to survive has also developed

3:34

quite a system of bureaucrats, and

3:36

they all want

3:38

to weigh in on Google's plans. So anyway,

3:40

I think another great podcast for our listeners

3:43

and a picture

3:46

of the week that's kind of a hoot too.

3:48

Oh, good. Always enjoy the pictures of the week.

3:51

Well, SecurityNow is ready to get underway. I

3:53

hope you are as well, boys and girls,

3:55

cats and kittens, club

3:57

members, and others. Now,

4:00

you club members, you don't have to hear this because

4:02

it's time for a commercial. Our

4:04

show today brought to you by, you

4:07

know the name, Melissa, the data quality

4:09

experts. Since 1985,

4:13

did you know Melissa's global

4:15

address verification and validation service,

4:18

I mean when they say global, they mean it, works

4:20

in 240, 240 plus countries and territories. I

4:24

didn't even know there were that many. Now,

4:27

why is that important? Well, no matter where

4:29

your business is located, now you can improve

4:31

deliverability worldwide, reduce costly

4:33

errors, you can boost your

4:35

address data accuracy everywhere, increase

4:38

delivery speed, boost ROI.

4:40

It is ultimately all about the return

4:42

on investment. That's why Melissa offer free

4:45

trials, sample codes, flexible

4:47

pricing. They know your bottom

4:49

line is important to you. They also

4:52

know your data is important to you.

4:54

And Melissa secures your data with the

4:56

highest quality standards. They are FedRAMP authorized.

4:59

Now, that's a huge deal. Of course, it doesn't

5:01

matter if you're not a government agency, but everybody

5:03

benefits from that security level. It's the best you

5:05

can get. All

5:08

Melissa users benefit from that also,

5:10

of course. Melissa's solutions and

5:12

services are GDPR and CCPA compliant.

5:14

They mean SOC 2 and HIPAA

5:16

high trust standards. So your information

5:18

is absolutely secure with Melissa. But

5:20

they can do so much with

5:22

it to really improve your

5:24

customer data. And better

5:27

customer data means a better bottom

5:29

line. Download the free Melissa

5:31

Lookups apps if you want to try it.

5:33

They're on iOS and Android. They're absolutely free.

5:35

Melissa Lookups with an S. You

5:38

don't even have to sign up. You can validate

5:40

addresses and personal identities in the US or Canada.

5:43

You can check global phone numbers, global IP

5:45

address information and more. It's really a handy

5:47

little tool. Get started today. 1000

5:50

Records Clean for free. Just to give you

5:52

a sample of what Melissa can do on-prem

5:54

in the cloud as a SaaS application. They

5:56

have an API so you can build it

5:58

in your own software. Go to

6:00

melissa.com/twit to know more.

6:04

Melissa, melissa.com/

6:08

twit. And we thank them so much for

6:10

the support of the good work, the important work

6:12

that Steve's doing right here at

6:14

Security Now. Well, the

6:16

important work will be appearing

6:19

shortly. Hey, this is

6:21

important work. Do not knock this work. Now

6:25

we have a picture, our picture of the

6:27

week, from somewhere looks like in the U.S.

6:31

Southwest. There's

6:33

no signs of any telephone poles

6:36

or structures, so we're kind of

6:38

out in the desert somewhere. And

6:42

so one of the things that people want

6:45

is they want their cell phones to

6:47

work out in the middle

6:49

of nowhere. And actually this is a problem

6:52

I have with many movies these days which

6:54

seem to forget that it's necessary to have

6:56

a cell tower not too

6:58

far away from where your

7:01

cellular device is in order

7:03

for it to get any connection. We

7:05

see people wandering out in the middle

7:08

of literally nowhere and they're on the

7:10

phone, unless the writers don't want

7:12

them to be, in which case they're holding the phone up,

7:15

scanning around, trying to find a signal. Well,

7:19

the way we solve the problem

7:21

of people wanting cell phone

7:23

coverage wherever they are, yet

7:25

nobody wanting to despoil the

7:27

landscape as a means of

7:29

providing it, is we come

7:31

up with stealth cell

7:34

phone towers. And

7:36

I'm not sure how truly

7:38

stealthful this is because it looks

7:41

a little square to be a

7:44

cactus. But I gave

7:46

this picture, they captioned, oh don't

7:49

mind us, we're just putting the lid

7:51

back on the cactus. Because

7:54

this is clearly a cell phone

7:56

tower cactus Which is meant

7:59

to... Don't read it

8:01

actually. if ya, it's got a little

8:03

extra. I wouldn't call

8:05

it arm off the side of the

8:07

cactus. Good looking, a little more to

8:10

make of whole thing, looking for more

8:12

tact the slightest As as said so

8:14

and actually you could see some cat

8:16

died in the neighborhood that look decidedly

8:18

less mechanical. They're not stay this way

8:20

all over Mexico, the city so are

8:22

of cactuses and I guess the southwest

8:25

as well. So be it are you

8:27

see it with a hundred others you

8:29

probably wouldn't look twice. Sexy.

8:31

However with sir it's certainly not Bruno

8:33

an eyesore looking like this thing would

8:35

look like at with the lid off

8:38

would I can see or because they

8:40

did it was a success at it.

8:42

At a crazy either billet is often

8:44

the crane is has is lifted the

8:46

yeah the lid off the cactus Anyway

8:49

I just got a kick out of

8:51

this just as and I I've seen

8:53

fake palm trees and I know that

8:55

here in on on the so called

8:57

the Sword now the famous four or

8:59

five in Southern California. Are

9:01

there are. Power. Lines

9:04

that run alongside the freeway

9:06

and. And. Every

9:08

like very often, there's a

9:11

big cluster of of of

9:13

cells. Equipment on these power

9:15

lines because it's a perfect place for

9:18

them to be. You know it's over,

9:20

there's already or I have a right

9:22

away there's saw ability to to to

9:24

run a service vehicle along the back

9:26

and and so forth and I many

9:28

many many moons ago it's back in

9:30

the spin Right actually was actress been

9:32

right to because I remember I was

9:34

working on Spin Right Three I built

9:36

a building. When. A lease.

9:39

So the a whole and I go

9:41

to a corporate headquarters. or twenty thousand

9:43

square feet, two stories and one point

9:45

so more my acres of land. That

9:47

so for now anyway. Ah, the cell.

9:50

Companies. came to me and

9:52

said hey ah this building is like

9:55

up on a point on a bluff

9:57

look at out over this valley you

10:00

can make some extra money by

10:03

letting us put some self things

10:06

along you know I'd like ringing along the edge

10:08

of your roof well you

10:10

know what my answer was you said

10:12

no answer you know this is a

10:14

beautiful building I'm not gonna have it

10:17

you know warts of self

10:19

crap all over that they're there now

10:22

Steve they are I

10:25

mentioned that because I drove by

10:28

not long ago you know looking

10:30

wistfully up at the building and

10:32

there it was just I don't

10:34

know I don't think you could

10:36

you could get more cell tower

10:38

crap around the perimeter of

10:40

this roof than there is there now

10:42

but not while I was in

10:44

control but immediately after I left

10:47

here anyway such

10:49

as the world you know and that's why I also have

10:51

no ads on my site I've Mark Thompson

10:53

made a case of what point he

10:55

said Steve there's something wrong

10:57

now with a website that doesn't have

11:00

a yeah what's wrong with you yeah

11:02

no thank you

11:06

anyway I wanted to

11:08

start off this week by sharing an important

11:12

piece of interesting news that's not

11:14

internet security related but is

11:16

nevertheless potentially quite a

11:18

big and serious issue in the

11:20

real world last

11:23

Thursday's headline and wired was

11:26

the dangerous rise in

11:28

GPS attacks with the

11:30

subhead thousands of planes and

11:32

ships are facing GPS jamming

11:34

and spoofing experts

11:37

are warning these attacks

11:39

could potentially impact critical

11:41

infrastructure communication networks and

11:44

more okay so I thought

11:46

that was interesting got my attention they

11:49

said the disruption to GPS services

11:51

started getting worse on Christmas

11:53

Day meaning at the end of 2023 planes and

11:55

ships moving around southern Sweden.

12:00

And Poland loss connectivity.

12:02

As. The radio signals were interfered with.

12:05

Since. Then the region

12:07

around the Baltic Sea,

12:09

including neighboring Germany, Finland,

12:12

Estonia, Latvia, and Lithuania

12:14

has faced persistent attacks

12:16

against Gps systems. Tens

12:19

of thousands of planes flying

12:21

in the region of reported

12:23

problems with their navigation systems

12:25

in recent months amid widespread

12:27

giant jamming attacks which make

12:29

Gps inoperable as the attacks

12:31

have grown. No. Surprise

12:34

to anyone. Russia. Has

12:36

increasingly been blamed with

12:38

open source researchers tracking

12:40

the source to Russian

12:42

regions such as Kaliningrad,

12:44

In. One instance, signals were

12:47

disrupted for forty seven hours

12:49

continuously. On. Monday murky

12:51

one of the most serious

12:53

incidents yet airlines thin Air

12:55

council it's flights to Tart

12:57

to Estonia for a month.

13:00

After Gps interference forced to have

13:02

it's planes to abort their landings

13:04

at the airport and turn around.

13:06

Part. About dependence on Gps apparently

13:09

uses can't land anymore without it.

13:12

The jamming in the Baltic region they

13:14

wrote which was first spotted an early

13:16

Twenty Twenty Two is just the tip

13:18

of the iceberg. In. Recent years,

13:20

there's been a rapid uptick

13:22

in attacks against Gps signals

13:24

and wider satellite navigation systems

13:27

known as G and Ss

13:29

as jill generic satellite navigation

13:31

including those of Europe, China,

13:33

and Russia. The attacks could

13:35

jam signals, essentially forcing them

13:37

off line or spoof the

13:39

signals, making aircraft and ships

13:42

appear as false locations on

13:44

maps, which you can imagine

13:46

my be even more damaging

13:48

than just jamming outright. Beyond

13:51

the Baltics war zone areas around

13:53

Ukraine and the Middle East have

13:55

also seen sharp rises and Gps

13:57

disruptions including signal blocking then to

13:59

disrupt airborne attacks which actually as

14:01

will see lil bit later the

14:04

I think is is the did.

14:06

The. Actual goal of this? because of

14:08

that the degree to which drones

14:10

are now using Gps. Wired.

14:13

Wrote now government's tell com

14:15

and airline safety experts are

14:17

increasingly sounding the alarm about

14:20

the disruption have the potential

14:22

for major disasters. Foreign ministers

14:24

in Estonia, Latvia, and Lithuania

14:26

of all blamed Russia. For.

14:29

Gps issues in the Baltics this

14:31

week and said the threat should

14:33

be taken seriously. Jamie. Adamson,

14:35

the Chief of public affairs for

14:37

the Swedish Navy, told Wired quotes

14:39

It cannot be ruled out that

14:41

this jamming as a form of

14:43

a hybrid warfare. With. The

14:45

aim of creating uncertainty and

14:47

unrest. Of course, there

14:49

are concerns mostly for civilian

14:51

shipping and aviation that an

14:54

accident will occur, creating an

14:56

environmental disaster. There's. Also a

14:58

risk that ships and aircraft will

15:00

suspend their traffic to this area

15:03

and thereby effect global trade. Your.

15:05

Wagner. A spokesperson from Germany's

15:07

Federal Officer Information Security told

15:10

Wired a growing threat situation

15:12

must be expected in connection

15:14

with Gps jamming. Wagner.

15:17

Said there are technical ways to

15:19

reduce its impact. Officials and villain

15:21

say they've also seen an increase

15:24

in airline disruptions in and around

15:26

the country. And a spokesperson for

15:28

the International Telecommunication Union, a United

15:31

Nations agency, told Wired that the

15:33

number of jamming and spoofing incidents

15:35

have increased significantly. Over. The

15:37

past four years. And. Interfering with

15:40

radio signals is prohibited under the

15:42

I to Use rules. g. using.

15:45

Russia has slowed down by

15:47

a Nato agency the International

15:50

Telecommunication Union saying well, you

15:52

should be doing math. right?

15:55

Attacks against Gps and the

15:58

wider Gnss category. In

16:00

two forms. First, Gps jamming

16:02

overwhelms a radio signals that

16:04

make up Cbs and make

16:07

the systems unusable. Second, spoofing

16:09

attacks, which actually are far

16:11

more sophisticated. Ten replaced the

16:13

original signal with a new

16:15

location Spoofed ships can, for

16:18

example, appear on maps as

16:20

if they're at inland airports,

16:22

and actually, that did happen

16:24

recently. Both types of interference

16:27

have increased in frequency disruptions

16:29

at least. At this stage

16:31

mostly impact planes flying at high

16:33

altitudes and ships. They can be

16:35

an open water not people's individual

16:38

Phones are other systems that rely

16:40

on Gps within the Baltic region.

16:43

Forty. Six thousand aircraft

16:45

showed potential signs of

16:47

jamming between August Twenty

16:49

Twenty Three and March

16:51

this year. According to

16:53

reports and data from

16:55

tracking service, Gps jam

16:57

Ben was Seagate. And.

16:59

An academic at the Zurich University

17:02

of Applied Sciences who also runs

17:04

alive Gps spoofing map there is

17:06

such a thing says there had

17:08

been an additional forty four thousand

17:11

spoofing incidents logged since the start

17:13

of this year. Earlier.

17:15

This month, more than fifteen

17:17

thousand planes earlier this month.

17:20

More. Than fifteen thousand planes

17:22

had their locations spoofed to

17:25

be rude Airport. According.

17:27

To data that Sergei shared

17:29

with Wired, more than ten

17:31

thousand were spoof to the

17:34

Cairo Airport, while more than

17:36

two thousand had their locations

17:38

showing in Yarrow Slavic Russia.

17:40

The data shows. Separate.

17:43

Analysis from Geospatial Intelligence Company

17:45

Geo Collect shared with Wired

17:48

showed that on April sixteenth,

17:50

around fifty five ships. Broadcasts.

17:53

Their location as being or

17:55

hover the main runway. At.

17:59

Us. Semper

18:02

All Paul International Airport in Crimea,

18:04

Ukraine. The airport is around ninety

18:06

miles inland from the Black Sea.

18:09

Words Believe the ships were actually

18:11

located. So yeah, it's no longer

18:13

possible to the believe what Gps

18:15

is showing you. You need to

18:18

look out the window and see

18:20

where you actually are. Zoc

18:23

Clements, a graduate research assistant at

18:26

the University of Texas here in

18:28

Austin, said the biggest change in

18:30

the past six months is definitely

18:33

the amount of spoofing. As I

18:35

said, spoofing as far more sophisticated

18:38

and difficult than just jamming and

18:40

potentially far more dangerous. He said

18:42

for the first time, we're seeing

18:45

widespread disruptions in civil aviation, especially

18:47

in the Eastern Mediterranean, the Baltics

18:50

and the Middle East in prior

18:52

years. There were reports of spoofing

18:54

impact the marine vessels. But. Not

18:57

aviation. Clemens. Says there

18:59

appear to be three spoofers that can

19:01

be traced back to Russia. One.

19:03

Open Source Intelligence L is going

19:06

by the pseudonym Marcus Johnson has

19:08

located jamming in the Baltics. At

19:11

which and that which impacted the

19:13

finish airline this week. So that

19:16

was the one that that was

19:18

the causing them troubled to Kaliningrad

19:20

of and other Russian locations. Research

19:23

group has suggested disruption near Poland

19:25

impacted Russia's own Gnss system less

19:27

than others. Not surprisingly, Russia doesn't

19:30

want to hurt themselves. That is

19:32

one of. A. Disturb everybody

19:34

else. And. Russia

19:36

has a long history of interfering

19:39

with Gps signals are both was

19:41

in Us borders and internationally Russia's

19:43

embassy, not surprisingly, in the Uk.

19:46

Did. Not respond to a request for

19:48

comment. The. disruptions can

19:50

cause uncertainty and potential safety

19:53

issues for airline pilots and

19:55

their passengers doubt no kidding

19:57

a spokesperson for euro control

20:00

European Aviation Organization with more than

20:02

40 countries as members

20:04

says its analysis shows disruptions

20:07

are happening in the eastern

20:09

Mediterranean areas around Ukraine

20:11

and the Black Sea as well as the

20:13

Baltic states during one week in

20:15

March four

20:18

thousand three hundred and eighty

20:20

seven aircraft reported issues the

20:23

euro control spokesman says for the same

20:25

time last week there were 2646 flights

20:29

reporting problems the

20:32

euro control spokesman says planes

20:34

can fly safely without GNSS

20:37

but interference quote puts

20:39

a higher workload on pilots and

20:41

air traffic control a safety

20:44

note is issued by the UK Civil Aviation

20:47

Authority this month says loss

20:49

of GNSS which is

20:51

just you know general satellite based navigation

20:54

can result in serious

20:57

navigation issues incorrect emergency terrain

20:59

warnings that the plane is

21:01

too low to the ground

21:03

and failure to various other

21:06

systems and

21:08

finally in a NASA report

21:10

detailing GPS instance there was also

21:12

published this month one pilot said

21:14

I have flown with

21:17

crew members who were not fully

21:19

aware of this problem other

21:21

pilots said they'd received false terrain

21:23

warnings that caused them to pull

21:26

up and the pilot

21:28

yes and the pilot should have

21:30

a thorough review of jamming effects

21:32

on the different aircraft systems as

21:34

part of their training and then

21:36

here's the problem of course because

21:39

this is a relatively new phenomenon

21:41

relative to when the pilots were

21:43

trained it may

21:47

just be the fact that the pilots are

21:49

trusting their avionics and no

21:51

not being sufficiently skeptical

21:54

so it does look

21:56

like these GPS

21:58

disruptions are coinciding with

22:00

Russia's full-scale war in Ukraine

22:03

and also it looks like

22:06

Israel's attacks in Gaza have

22:09

also been tied into

22:11

this. As we know,

22:13

disrupting GPS as part of

22:15

electronic warfare has become

22:18

common on Russia and

22:20

Ukraine's battlefield as a way to try

22:22

to limit the operation of drones. And

22:25

while Iran launched a barrage of missiles

22:27

and drones against Israel last

22:29

month on the 13th, Israeli GPS

22:32

disruption designed to limit the impact

22:34

of the attack also impacted mapping

22:36

and taxi services as well as

22:39

food delivery. So here was

22:41

an instance of Israel

22:43

doing some GPS jamming

22:45

which was somewhat

22:47

indiscriminate and the mapping

22:49

and taxi services as well as food

22:52

delivery within their own country took

22:54

a hit as a consequence. Kevin

22:57

Henke writes Wired, the founder

22:59

of cybersecurity company Hensec whose

23:01

work includes detecting GPS disruption

23:04

says jamming and spoofing technology

23:06

has become cheaper and

23:08

smaller over the years to the extent

23:11

that individuals can install them in their

23:13

cars to hide their own movements. That

23:15

is, you know, you're blocking your own

23:17

GPS receiver so your car doesn't know

23:20

where it is. However,

23:22

Henke says more sophisticated attacks use

23:24

equipment that can cost huge sums.

23:26

Yes, anything you anytime you're doing

23:28

spoofing as I said spoofing is

23:30

is a whole other level than

23:32

than just blanket jamming. He

23:35

said in conflict zones in military

23:37

terms and in professional terms this

23:39

spoofing is very sophisticated and it

23:41

always goes hand in hand with

23:44

jamming. Okay,

23:46

so since both

23:48

the jamming and the location spoofing

23:51

disruptions are enabled through the

23:53

use of very powerful

23:55

Local radio transmitters which over

23:57

help which overwhelm the reception.

24:00

The of the authentic signals

24:02

being beamed down from the

24:04

Gps satellites systems in orbit.

24:07

So. long as you're not in the

24:09

region, Of. The Baltics,

24:11

And all were Russia appears to have

24:14

taken it off. Serious.

24:16

Action To create Major disruptions.

24:19

The good news is these

24:21

attacks are inherently local in

24:23

nature and I'll. Hear the

24:25

U S. Were. Not being

24:27

affected by it at all,

24:29

as as most of Europe,

24:31

they are inherently very local.

24:33

But the problem for those

24:35

who are in the region

24:37

is it Gps and the

24:39

wider Gnss which again global

24:41

Navigation satellite system have always

24:44

been. Incredibly. Reliable

24:46

sources and not just of

24:48

of location but also of

24:51

time they they are masters

24:53

sources of of especially time

24:56

of day and as we

24:58

know when something is both.

25:01

Very. Useful. And. Has earned

25:03

a reputation for also being very reliable.

25:06

I mean, you know these things are

25:08

up in the sky beaming down at

25:10

us? Did. They. End up

25:12

creating a strong dependence. We

25:14

end up becoming very dependent

25:16

upon them. So many. Modern

25:19

non military commercial systems

25:21

have become so reliant

25:23

upon Gps that the

25:25

deliberate disruption of that

25:27

service for military purposes

25:29

such as Russia, as

25:31

has likely been perpetrating,

25:33

can cause dramatic collateral

25:35

damage. The.

25:38

us that day the gps system

25:40

which is i put out

25:42

by the u was conceived quite

25:44

a while ago a little over

25:47

fifty years ago back in

25:49

nineteen seventy three it took five

25:51

years to package this in the

25:54

first satellite that began launching at

25:56

that time and today we have

25:58

twenty four satellites in up

26:01

in the GPS constellation. They've

26:03

been up and operating since 1993. And

26:08

talk about depending upon something that's more

26:10

fragile than we might want. Our

26:13

phones and automobiles today only know

26:15

where they are largely thanks to

26:17

GPS signals coming from space. We've

26:19

talked about the mind. I only

26:21

know where I am thanks to

26:23

GPS signals. Yeah. Forget the car.

26:27

Yeah. I can't drive without GPS.

26:29

And I'm sure sports

26:32

wristwatch, health tracking wristwatches

26:35

are doing the same thing. We

26:39

have recently been talking about the

26:41

militarization of space and

26:44

the idea that having satellites attacking

26:46

one another up there

26:48

is not a

26:50

worry of James Bond science fiction.

26:53

You know, it's actually happening.

26:56

In some cases, robot

26:58

satellites are there in order to

27:00

repair others. But the same robot

27:02

that can function, you know, to

27:04

fix a broken antenna can also

27:06

go over and break one off

27:08

of some other satellite. So,

27:11

you know, unfortunately, they also have

27:13

multiple purposes. And

27:15

unfortunately, as global political tensions increase, we

27:17

can hope and we need to hope

27:20

that no major powers having

27:23

space based military capabilities, nor

27:25

the ability to kill satellites

27:27

from the ground, believe

27:29

that denying the entire world these benefits

27:32

would create an advantage for them. Because

27:35

it's difficult now to conceive

27:37

of a world where, you

27:39

know, GPS was just shut

27:42

down. It was like

27:44

destroyed deliberately by a power

27:47

hostile to, you know,

27:50

it wouldn't even necessarily have

27:52

to be hostile to the US. It

27:54

could be, you Know, because everyone's

27:57

using GPS, killing it for.

28:00

For Air Air, everyone also succeed

28:02

in killing it for a specific

28:04

targeted country. Before.

28:06

Gps. The. Only way

28:09

for something you know where it was.

28:11

Was. Through a system of

28:13

inertial navigation, Inertial.

28:15

Navigation, like his name suggests,

28:18

is a closed system. Which.

28:21

Relies upon the systems

28:23

precise measurement of it's

28:25

own linear and angular

28:28

accelerations. It integrates

28:30

those over time to

28:32

determine it's velocities. And.

28:34

That integrates those overtime to

28:37

determine his position. Even.

28:39

Though inertial navigation systems are

28:41

still in use due to

28:43

the nearly instantaneous position and

28:46

especially angular feedback that they

28:48

provide, The. Errors that tend

28:50

to creep in overtime can only

28:53

be eliminated with the use of

28:55

slower but far more accurate input

28:57

from the global Gps system. I

29:01

suspect Russia's primary concern is

29:03

with the use of autonomous

29:05

military drones which may rely

29:08

upon Gps to determine their

29:10

in location. But since the

29:12

risks presented by Gps jamming.

29:15

Although. They have been prevalent

29:17

at it has a been

29:19

a big concern for airline

29:22

pilots until recently Oscars they

29:24

are operating over there. The

29:26

East in the Baltic areas

29:28

says jamming has been a

29:30

possibility for some time. I

29:32

suspect that the latest technologies

29:34

are much more immune to

29:36

Gps outages than those in

29:39

Russia might wish. Vivid all

29:41

of the advantages and the

29:43

advantage of the advances made

29:45

in vision. and in

29:47

real time recognition i would

29:49

be surprised if the latest

29:51

autonomous technologies were not able

29:54

to fly nearly as well

29:56

by sight as they can

29:58

these days by gps They

30:00

might well use GPS as a first

30:03

choice, but use vision

30:05

to detect location spoofing while

30:08

also being able to switch to

30:10

pure vision if GPS should fail

30:12

completely. And another likely

30:14

strategy, which again you don't worry about

30:17

or deal with until it becomes a

30:19

problem, is that

30:21

since GPS signals will

30:23

always be originating from

30:25

above, would be to shield any

30:27

GPS receiver and its antennas. Oh, from

30:29

below. Yes. Because

30:32

the jammers are on the ground. Exactly.

30:35

That's clever. Yeah. So

30:37

planes can do that because they're

30:39

well above ground. Unfortunately, it's probably

30:41

not practical for ships at sea.

30:45

Yeah. I mean, when you listen to ground

30:47

air traffic control, talking to an airplane, which I

30:49

used to always do on United Channel 9, I

30:52

used to love to do that. They

30:54

often have visual markers.

30:57

They say turn right at

30:59

the big rock candy

31:01

mountain and things like that. I

31:04

don't know if they still do that. I haven't

31:06

listened in a while, but I bet they do.

31:10

You always want redundancy in any

31:12

system like that, right? Yeah.

31:15

And of course, the problem is that we all...

31:20

Okay. I remember,

31:23

Leo, when I guess this must

31:25

have been in driver ed. We were

31:27

supposed to go out

31:30

and walk around our car to

31:32

check all four tires. Yeah. We

31:34

don't do that anymore. No. Do

31:37

you do that? When was the last time

31:39

anybody did that? Pilots do that. Commercial

31:41

jet pilots do that. And I thank goodness that

31:43

they do. I think that's really great. But

31:46

no, I haven't done that to my car in a while. I figure if

31:48

it's flat, I'll know. Right?

31:51

It'll go from... That's right. From...

31:53

That's right. But I do remember

31:55

being told that's what we're supposed to do. So

31:58

here we have a problem where... G

32:02

been so reliable and

32:04

relied on that that

32:06

I'm just hoping I mean the

32:09

in this NASA report last week where one

32:11

of the guys said you know I've

32:13

been with flight crews that

32:15

just assumed that the GPS was telling

32:18

the truth even though they were

32:20

suddenly being told to pull up because you're about

32:22

to hit the rock candy mountain that

32:24

would not be good. Pull

32:28

up, pull up. Take another break

32:30

and then we're going to talk about

32:32

whether the sky is falling on all

32:34

VPN systems. Yeah. As the tech press

32:37

seems to believe. I was counting on

32:39

you to cover this because I read

32:41

the stories. Thank God

32:43

you're covering it before I actually did

32:45

the stories. Keep me out

32:47

of trouble. Please. We

32:49

are very proud to say that our show this

32:52

portion of the show brought to you by

32:54

collide. We're going to meet with collide tomorrow

32:56

at RSA. Great people with a great product.

33:00

In fact I was so happy when I heard

33:02

and maybe you heard too that collide was just

33:04

purchased by one password. That's

33:06

really good news. Two

33:08

companies leading the industry

33:10

in creating security solutions that put

33:12

users first. They belong together. For

33:14

over a year we've been telling

33:16

you about collide device trust helping

33:18

companies that use Okta make sure

33:20

that only known and secured devices

33:23

can access the data.

33:26

Okta authenticates the human.

33:30

Collide authenticates the hardware. That's

33:32

a big deal and

33:34

they're going to still doing that as part of one

33:36

password. In fact they're going to even be better at

33:38

this. If you've got Okta and you've been meaning to

33:40

check out collide this is the

33:42

time. Absolutely collide is easy to get

33:44

started with. They come with a pre-built

33:46

library, a pre-built device posture check so

33:48

you can get up and running right

33:50

away all the basic stuff you would

33:52

want. But if you have some specific

33:54

devices or software or situations you'd like

33:56

to check for it's very easy to

33:58

write your own custom. checks for just

34:01

about anything you can think of. Another point

34:03

I really like, you

34:05

can use Collide on devices without MDM.

34:08

That means your Linux fleet or your

34:10

contractor devices, you can't tell them, put

34:12

our MDM on your device and

34:15

of course every BYOD phone and laptop

34:17

that sneaks into the company, all

34:19

of them can be protected by Collide. Now

34:22

that Collide is part of OnePassword,

34:24

it's only going to get better.

34:27

This is the time to check

34:29

it out. collide.com/security now. Learn more,

34:31

watch the demo today. k-o-l-i-d-e.com/security now

34:33

and if you're at RSA go

34:36

visit them. They're at us

34:38

RSA too. Starting, I think it starts

34:40

tomorrow in San Francisco. Collide.

34:42

Thank you Collide for supporting

34:44

Steve and his very important

34:47

mission. Now what's all

34:49

this about VPN Steve? Okay,

34:52

so yesterday

34:55

ours Technica got a little carried away

34:57

in their reporting of

35:00

what amounts to a clever hack that

35:02

a Seattle Washington based pen and

35:05

pen testing firm known

35:07

as the Leviathan security

35:09

group posted in their

35:11

blog and of course the rest

35:13

of the tech press picked up on it quickly too.

35:16

The blog posting carried the

35:18

headline how attackers

35:21

can decloak routing based

35:23

VPNs for a total

35:25

VPN leak and

35:27

what I found curious

35:29

was that they assigned, they meaning

35:32

the Leviathan security group, assigned a

35:34

CVE number to their discovery

35:36

even though nothing about

35:39

this is a bug or a

35:41

flaw. Oh. It's just a clever

35:43

local exploit of a little used

35:45

feature of DHCP servers. Unfortunately,

35:48

ours Technica's

35:51

headline for their story was novel

35:54

headlined novel attack

35:56

against virtually all VPN

35:59

apps neuters their

36:01

entire purpose. Ahhh!

36:05

Run away! Okay,

36:08

which of course makes this sound more like

36:10

the end of VPNs as we've known them.

36:13

It isn't. Here's what's going on. Okay,

36:16

so I'm going to do a bunch

36:19

of propeller head cool stuff in order

36:21

to get a real grip

36:23

on this. Our PCs

36:26

all interact with both

36:28

internal and external networks

36:30

through network interfaces. Most

36:33

systems typically have a single physical network

36:35

interface or NIC, but it's

36:37

possible for a machine to have

36:39

more than one physical network interface

36:41

with each interface connected to a

36:44

different physical network. In

36:46

that case, it's important for

36:48

outgoing network traffic to

36:51

know which physical interface any

36:54

given packet should be routed out through.

36:58

To answer that question, our machines

37:00

contain a routing table. The

37:03

routing table performs a

37:05

most specific match function

37:07

based upon the destination IP address.

37:10

And in years past we talked

37:12

about internet routing tables and

37:14

all of this, so we've covered this

37:16

in detail, but the key here is

37:19

most specific match and that

37:21

all of our PCs, every one of them,

37:24

pads, phones, you name it,

37:26

anything that's networked using internet

37:28

protocol, IP protocol has a

37:30

routing table. Under

37:35

Windows, for example, opening a

37:37

command prompt and entering

37:39

the command route space

37:41

print will display a

37:44

list of the system's

37:46

interfaces followed by the

37:48

IPv4 and IPv6 routing

37:51

tables respectively. And

37:53

they're interesting and you can get a

37:55

sense for the fact that there's a

37:57

lot going on under the covers. that

38:00

we don't appreciate, we normally don't even

38:02

see. Okay, so

38:05

this set of network communication,

38:08

that is IP-based network communication,

38:10

comes in so handy that

38:13

in addition to true physical

38:15

interfaces, many of

38:17

our machines will have one or

38:19

more virtual network interfaces. In fact,

38:22

the so-called local host,

38:25

127.0.0.1, that's

38:28

a virtual network interface that

38:30

all stacks have. And

38:33

for example, the use of virtual

38:35

machines has become very popular and

38:37

they create their own virtual network

38:39

interfaces to talk to their host

38:41

machine as well as to the

38:44

outside world. Okay,

38:47

so here's the main point. Many

38:50

VPNs, like OpenVPN

38:52

for example, operate

38:56

by creating their own virtual

38:58

interface in the hosting machine.

39:01

It looks like and operates

39:04

like any other network interface.

39:07

But being a VPN, a virtual

39:09

private network, which is used to

39:11

transact privately with encryption,

39:14

any packets sent out of

39:17

that virtual interface are

39:19

first encrypted, then rerouted

39:22

out of an actual

39:24

physical interface to be

39:26

sent to the VPN's matching endpoint.

39:29

Since the typical VPN user, while

39:32

using a VPN, wants all of their

39:34

machines' traffic to be tunneled through

39:37

the VPN, when

39:39

the VPN tunnel is brought up, the

39:42

VPN software dynamically

39:44

edits the system's

39:46

global routing table in

39:49

such a way that instead of

39:51

the system's traffic by

39:54

default being routed out

39:56

through its normal actual

39:58

physical interface, All

40:00

of its traffic is instead routed

40:03

to the VPN's software-created

40:05

virtual network interface. This

40:08

is the way that deep down inside the

40:10

guts of our machines, all of

40:13

the traffic that is normally unencrypted

40:15

suddenly becomes encrypted when we

40:18

activate our VPN. Essentially,

40:20

it's like a man in the middle.

40:22

It sticks a shim into our network

40:24

so that all of the traffic that

40:26

would normally just go straight out the

40:29

physical interface instead is routed

40:32

to the VPN. And that's done,

40:34

as I said, by making just

40:37

a slight change to the routing

40:39

table so that all of

40:41

the traffic instead of going out the physical

40:43

interface goes to the VPN. We

40:46

need one other piece of information just to be

40:48

certain that everyone's on the same page. DHCP

40:53

stands for Dynamic Host

40:55

Configuration Protocol. By

40:57

default, when any networked

40:59

machine boots up and gets itself

41:02

going, it needs to be

41:04

using an IP address for itself on

41:06

its local network that's unique for

41:08

that network. And it

41:10

needs to know the IP address to which

41:13

it should address packets bound for the outside

41:15

world. In other words, the

41:17

network's gateway IP. It

41:19

may also want to know the IP

41:21

addresses of some DNS servers that will

41:24

honor its requests for domain

41:26

name lookup. It's

41:29

the network's inward-facing

41:32

DHCP server that

41:34

answers all these needs. When

41:37

any networked machine starts up, by

41:39

default, it will emit a broadcast

41:41

packet onto the network, announcing its

41:44

presence and asking for

41:46

any listening DHCP server to please

41:48

provide it with all the information

41:50

it requires to become a well-behaving

41:52

citizen on the local network and

41:55

to connect to the rest of

41:57

the global internet. DHCP

42:01

cleanly organizes the various types

42:03

of information it can supply

42:05

into like to the clients

42:09

who are requesting it by

42:11

number. Each

42:14

one of these is known as an option where

42:17

the option number is a single byte,

42:20

thus having a value from 0 to 255. 0

42:24

is a null option and it can be used for

42:26

padding. 255

42:28

is the marker for the end of the

42:30

list of options. So the

42:33

options are provided as a list

42:35

of information terminated by option

42:37

255, which of course you

42:40

know is a byte of all 1s. So

42:43

for example, option 1

42:45

provides the network's subnet mask

42:48

to the requesting client. Option

42:51

2 specifies the offset of

42:53

the client's subnet in

42:56

seconds, that is in real

42:58

time, from UTC,

43:01

Coordinated Universal Time. Option

43:04

3 specifies a list of

43:06

the IP addresses of routers

43:08

on the client's subnet, what we know

43:10

as the gateway IP. Option

43:13

4 specifies a list of time servers

43:15

which are available to the client. Option

43:18

6 provides a list of DNS servers

43:20

for the client's use. And

43:23

there's a bunch of them, all

43:26

kinds of different things that have been added

43:28

through the years and there are even some

43:30

surprises. For example, options 69 and

43:33

70 provide the IP addresses

43:35

of SMTP and POP3 email

43:38

servers, which I thought was

43:40

kind of cool. We're

43:42

all used to specifying those ourselves, but back in

43:44

1997 when this was first created, that

43:49

information was available via

43:51

DHCP. Something

43:54

else that DHCP was able

43:56

to provide is the

43:58

source of today's webinar. trouble.

44:01

The RFC's definition for

44:04

option 33 defines

44:07

it as the static

44:09

route option and

44:11

says quote this option

44:14

specifies a list of static

44:16

routes that the client should

44:18

install in its routing

44:21

cache. Okay now everybody

44:23

who's been paying attention and

44:25

you know enjoys networking stuff

44:28

just went aha and knows

44:30

what the problem is. This

44:33

thing continues if multiple routes to

44:35

the same destination are specified they're

44:37

listed in descending order of priority.

44:40

The routes consist of a list

44:42

of IP address pairs. The first

44:44

address is the destination address and

44:46

the second address is the router

44:49

for the destination. Again

44:51

if some of you just said oh crap

44:54

that would be the correct reaction.

44:57

What this means and

44:59

it would mean you're paying attention good job.

45:01

That's right that's right. What

45:04

this means is that the

45:06

response from a DHCP server

45:09

can be used to mess

45:11

with a machine's routing table

45:15

and as we noted earlier a machine's traffic

45:18

is routed to the

45:20

VPN's virtual interface through

45:22

a dynamic modification of

45:25

the machine's routing table. Now

45:28

as it happens option 33 is not really the problem

45:32

because it was defined back in 1997 when IP networks

45:35

were all class A, B, or

45:41

C. That meant that

45:43

networks were defined to always have

45:45

exactly one, two, or

45:48

three bytes of host

45:50

machine addresses. As

45:52

we know this was extremely

45:54

wasteful of IP addresses for

45:56

networks falling into intermediate sizes.

45:59

So something known as CIDR,

46:01

which stood for classless

46:07

inter-domain routing was adopted. That's what we

46:09

have today where the network mask can

46:11

have any number of contiguous bits set,

46:13

thus allowing scaling of networks by factors

46:16

of two all the way

46:18

from one machine, well technically up

46:20

to 4.3 billion, but no one

46:22

network has that except the internet

46:24

itself. Okay,

46:27

so the adoption of

46:29

CIDR obsoleted option

46:31

33 forcing its replacement five

46:33

years later in 2002 under

46:38

the guidance of RFC 3442 which introduced

46:40

option 121 which allows for exactly

46:47

the same thing but for

46:50

the but under the specification of

46:52

classless static routes. Now

46:56

I mentioned that I was

46:59

surprised that these LaVaya LaVaya

47:01

Thonsor security group guys had

47:03

arranged to get a CVE

47:06

assigned for this since technically

47:08

this is a feature not a

47:10

bug and all the way back in

47:12

1997 the

47:15

fundamental vulnerability of

47:18

DHCP was quite well

47:20

understood. Again 1997 section

47:24

7 of the original RFC

47:26

2131 dated March of 1997 is titled if it

47:28

was section

47:32

number 7 security considerations

47:35

it says DHCP

47:37

is built directly

47:39

on UDP and

47:41

IP which

47:43

are as yet inherently

47:46

insecure. Furthermore

47:48

DHCP is generally

47:50

intended to make maintenance of

47:53

remote and or diskless

47:55

hosts easier. While perhaps

47:57

not impossible configuring such

48:00

hosts with passwords or keys may

48:02

be difficult and inconvenient. Therefore,

48:05

DHCP in its

48:08

current form, which by the way is the

48:10

form it has today in 2024 because,

48:13

you know, if it's not broke, in

48:16

its current form is

48:18

quite insecure, says

48:20

the RFC from 1997. They

48:24

said unauthorized DHCP servers may

48:26

be easily set up. Such

48:29

servers can then send false

48:32

and potentially disruptive information to

48:34

clients such as incorrect or

48:36

duplicate IP addresses, incorrect

48:39

routing information, including

48:42

spoofing routers, etc.,

48:44

incorrect domain name server addresses

48:47

to spoof name servers,

48:49

and so on. Apparently

48:51

they wrote, once this seed

48:53

information is in place, an

48:55

attacker can further compromise affected

48:57

systems. Okay,

49:01

so here's how the

49:03

Leviathan folks described the

49:05

attack they've devised by

49:07

abusing option They

49:11

said, our

49:13

technique is to run

49:15

a DHCP server on

49:18

the same network as a targeted

49:20

VPN user and to

49:22

also set our DHCP configuration to

49:24

use itself as a gateway. When

49:28

the traffic hits our gateway, we

49:30

use traffic forwarding rules on the

49:33

DHCP server to pass traffic through

49:35

to a legitimate gateway while we

49:37

snoop on it. We

49:41

use DHCP option 121

49:44

to set a route on

49:46

the VPN user's routing table. The

49:49

route we set is arbitrary and

49:52

we can also set multiple routes if needed

49:55

by pushing routes that are more

49:57

specific than a slash zero signup.

50:00

range that most VPNs use,

50:03

we can make routing rules that have

50:05

a higher priority than

50:07

the routes for the virtual

50:10

interface the VPN creates. As

50:13

we know, because that

50:16

means it's a more specific

50:18

route, so the routing system

50:21

will always route a

50:23

more, will always take the

50:25

most specific route available. So

50:27

by doing something creating

50:30

a network smaller than the slash zero,

50:33

which is the everything, the

50:36

routing table ends up routing

50:38

to the intercepting

50:40

DHCP server rather than

50:43

to the users VPN.

50:46

They said we can set multiple

50:48

slash one routes to recreate the

50:50

zero dot zero dot zero slash

50:52

zero all traffic rule

50:55

set by most VPNs. Pushing

50:58

a route, they wrote, also means

51:00

that the network traffic will be

51:02

sent over the same interface as

51:04

the DHCP server instead

51:06

of the virtual network interface. This

51:09

is intended functionality that is

51:11

not clearly stated in the

51:13

RFC. Therefore, for

51:16

the routes we push, it

51:18

is never encrypted by

51:20

the VPNs virtual interface but

51:22

instead transmitted by the network

51:24

interface that is talking to

51:26

the DHCP server. As

51:29

an attacker, we can select which

51:31

IP addresses go over the tunnel

51:33

and which addresses go over the

51:35

network interface talking to our VPN

51:37

or our DHCP server. So in

51:39

other words, they're able to literally

51:42

select by destination IP.

51:44

If they don't want

51:46

everything, they can say, ah, just give

51:49

us this chunk of your traffic. You

51:51

think it's going through your VPN, but

51:53

it's not. They said

51:55

we now have traffic being

51:58

transmitted outside the VPN. encrypted

52:00

tunnel. This technique can

52:03

also be used against an

52:05

already established VPN connection once

52:07

the VPN users host needs

52:09

to renew a lease from

52:11

our DHCP server. We

52:14

can artificially create that scenario by setting

52:16

a short lease time in the

52:18

DHCP lease so the user updates

52:21

their routing table more frequently. In

52:23

addition the VPN control

52:25

channel is still intact

52:27

because it already uses

52:30

the physical interface for its

52:32

communication. That is you

52:34

know the control channel meaning the

52:37

channel to the remote end that is outside of

52:39

the tunnel. They said in

52:41

our testing the VPN always

52:44

continued to report as

52:47

connected and the kill switch

52:49

was never engaged to drop our

52:51

VPN connection meaning there was never

52:53

a panic that the VPN was

52:56

concerned that it was being intercepted

52:58

and so shut things down. So

53:02

then to their credit they raised

53:04

the question that we've

53:06

had all along by asking

53:08

is tunnel vision a vulnerability

53:11

and I appreciated their answer

53:13

they wrote this is debatable.

53:17

We're calling it a technique because

53:20

tunnel vision doesn't rely

53:22

on violating any security

53:24

properties of the underlying

53:26

technologies. From our perspective

53:28

tunnel vision is how

53:31

DHCP routing tables and

53:33

VPNs are intended to work.

53:36

However it contradicts

53:38

VPN providers assurances

53:42

that are commonly referenced

53:44

in marketing materials. In

53:46

our opinion tunnel vision becomes

53:48

a vulnerability when a VPN

53:51

provider makes assurances that their

53:53

product secures a customer from

53:55

an attacker on

53:57

an untrusted network. big

54:00

difference between protecting your data

54:02

in transit and protecting against

54:04

all LAN attacks. VPNs

54:07

were not designed to mitigate

54:09

LAN attacks on the physical

54:11

network and to promise otherwise

54:14

is dangerous. In

54:16

our technique, we have not

54:18

broken the VPN's cryptographically secured

54:20

protocol and the VPN is

54:22

still fully functional. An

54:24

attacker is instead forcing a

54:27

target user to not use

54:29

their VPN tunnel. Regardless

54:31

of whether we classify this as

54:34

a technique, VPN users are affected

54:36

when they rely on assurances that

54:38

a VPN can secure them from

54:41

attackers on their local

54:43

network. Hmm, interesting. Finally,

54:46

That is one of the primary uses, isn't it,

54:48

for a coffee shop and other open Wi-Fi networks?

54:51

Exactly. But that's been

54:54

around forever. Yes,

54:56

exactly. And

54:58

they finished. As for what systems are

55:01

affected, the short version is everything

55:03

except Android. That's funny.

55:08

Android doesn't support option 121, so

55:10

it's completely excluded from these attacks.

55:13

They wrote, In our

55:15

testing, we observed that any

55:18

operating system that implements a

55:20

DHCP client according to its

55:22

RFC specification and has support

55:24

for DHCP option 121 routes is affected. This

55:30

includes Winix, Windows,

55:34

Linux, iOS, and Mac OS. Notably,

55:41

they wrote, it does not affect

55:43

Android as they do not have

55:45

support for DHCP option I

55:49

wonder why not. Which really is interesting. I

55:52

do too because I did some

55:54

digging and there have actually been

55:56

instances where Android's lack of option

55:59

121 support has

56:01

caused problems for Android users

56:03

because it turns out this

56:05

is not obscure Leo

56:08

this is the first time we've ever talked

56:10

about it on the podcast because it's just

56:12

never come up you know we've covered DHCP

56:14

in depth in the past okay

56:16

so just to be clear about the

56:19

scope of the danger presented

56:21

by the potential abuse of

56:24

DHCPs option 121 this

56:26

is strictly a local

56:29

landside attack but Leo as you

56:32

correctly point out you know we

56:34

do operate in essentially

56:37

land networks

56:39

where we're assuming a VPN

56:41

is going to trust us

56:43

where untrusted peers

56:47

are on the same land we

56:49

are so that's

56:51

a thing the

56:57

attacker needs some means

56:59

of defeating the network's

57:01

actual DHCP server since

57:04

DHCP clients will and

57:06

do accept

57:09

the first reply to their

57:11

query simply being

57:13

a faster to reply is

57:16

typically all that's needed and you

57:18

know as we know most routers

57:20

use the slowest chip that the

57:22

manufacturer was able to get away

57:24

with boy I tell

57:26

you those web interfaces on routers it's

57:28

like okay I click the button hello

57:30

hello did

57:32

you you know try click it again or just wait

57:34

no so the point is it's not going to be

57:37

quick to fire off a

57:39

DHCP reply because it doesn't need

57:41

to write that's going to be

57:43

way down the priority queue of

57:45

traffic that it needs to deal

57:47

with so an attacker probably doesn't

57:50

have any choice have much difficulty

57:52

being able to respond with DHCP

57:54

queries faster so it's

57:56

definitely conceivable also

57:59

in an

58:01

enterprise environment that if

58:05

you had somebody untrusted on

58:07

an enterprise network that

58:10

would be a problem. And it also

58:12

turns out that option 121 is not the

58:15

least bit obscure in the enterprise.

58:17

It turns out it's under heavy

58:19

use. I found two little samples

58:22

through a quick search. A posting

58:24

over on Stack Exchange says, I'm

58:27

running OpenVPN on a

58:29

CentOS 7 server. The

58:32

DHCP server on the LAN

58:34

uses option 121 to tell

58:37

other devices to use

58:39

this CentOS server if they

58:42

want to get to the VPN

58:44

subnets the OpenVPN servers

58:46

connected to. This works

58:48

great. The problem is that

58:51

this CentOS server is getting these

58:53

same routes from the DHCP server

58:55

which breaks things. And then he

58:57

goes on to talk about how

58:59

he can manually remove the static

59:02

routes that the CentOS server is

59:04

receiving from DHCP. But my point

59:06

is here's an example of

59:09

where option 121 is being

59:12

used to inform

59:16

machines on the LAN

59:19

where to route the traffic they

59:21

want to go through the

59:24

CentOS 7 servers

59:27

VPN subnets. So

59:30

it's very useful for that. And

59:32

also just as recently as last

59:34

Tuesday someone posted to the what

59:37

I have to categorize as

59:40

the embarrassingly useless Microsoft

59:42

answers forum. I don't

59:45

know if anybody has ever seen any

59:47

of the crap that is there.

59:51

But you know if Microsoft

59:53

really wants to lead in AI

59:55

they should remove whatever

59:57

you whatever poor humans they have.

1:00:00

that are being forced to respond

1:00:02

to forum postings there and put

1:00:04

chat GPT 12 or

1:00:06

something in there instead it is I mean

1:00:08

it is it is excruciatingly

1:00:12

bad anyway Someone

1:00:14

posted and news will say they

1:00:16

got no useful answer when connected

1:00:18

to my office network It's

1:00:21

DHCP server meaning

1:00:23

his offices Network DHCP

1:00:26

server will use option

1:00:29

121 to assign three different networks

1:00:31

to be reached using a router

1:00:33

which is not the default gateway

1:00:37

This works absolutely the networks

1:00:39

appear in my routing table

1:00:41

in active routes everything works

1:00:43

Networks are reachable anyway So he

1:00:46

wrote that and I just grabbed that as a little

1:00:48

snippet of another example of like you know Option

1:00:50

121 is really out there and it

1:00:53

turns out has you know Really

1:00:56

been useful as I said he goes

1:00:58

on to explain in some length He's

1:01:00

complaining that when he boots his PC

1:01:02

without any network connectivity Then

1:01:05

it has a problem Yeah,

1:01:09

that would be a problem. So

1:01:11

anyway, I wanted to

1:01:13

point this out again that this

1:01:15

this this DHCP option is

1:01:19

In heavy use within more

1:01:21

complex corporate networks What

1:01:23

that means is that simply like? Blacklisting

1:01:26

option 121 is not viable in My

1:01:32

opinion it would be extremely unlikely

1:01:35

for anyone at home to ever

1:01:38

have anything You know to

1:01:40

worry about Though

1:01:42

it's still instructive to paint a picture the

1:01:45

way I can see this might occur

1:01:47

to somebody at home Would

1:01:49

be if some malicious device were

1:01:51

connected to a residential

1:01:53

network and wish to

1:01:55

capture all of the users Traffic whether

1:01:58

tunneled through a VPN or not By

1:02:00

being the first device to respond

1:02:04

to any DHCP query, such

1:02:06

a malicious device could establish

1:02:08

itself as the network's

1:02:11

gateway to receive, inspect, and

1:02:13

forward all traffic from

1:02:15

the network's many machines. And

1:02:18

then, by additionally using option

1:02:21

121, such a device could

1:02:23

use that to also insert

1:02:25

entries into the user's routing

1:02:27

table to prevent their VPN,

1:02:30

if any, from tunneling

1:02:32

the user's traffic. Even

1:02:34

though the VPN would show that

1:02:37

everything was working and the user's

1:02:39

traffic was protected, none

1:02:41

of it would be. The VPN tunnel

1:02:43

would be up and established, but it

1:02:45

would not be carrying any of the

1:02:48

user's traffic. Since there

1:02:50

are many environments where option 121 is not

1:02:52

needed and is never used, like probably

1:02:57

most of ours at home, I think

1:03:00

it would be nice for our operating

1:03:02

systems to provide the option to hard

1:03:07

disable it, but I

1:03:09

dug around that I couldn't find any

1:03:11

indication that that's being done. I

1:03:14

would imagine the Windows firewall could be

1:03:16

configured to

1:03:19

look for any incoming DHCP

1:03:21

port. What is it? It's

1:03:24

been so long. Is it 163? No

1:03:26

idea. The DHCP? I don't

1:03:29

remember now. The port number. The

1:03:32

best mitigation would be turn off option

1:03:35

121, but that's not an option for

1:03:37

us now. Can

1:03:40

VPN software be updated to have that as

1:03:42

a feature? The

1:03:46

problem is this gets in underneath

1:03:48

the VPN software. I

1:03:53

suppose it could be updated

1:03:55

to monitor the routing table

1:03:58

and proactive. determine

1:04:01

whether or not it's been

1:04:04

rerouted. So that's

1:04:06

certainly something that could be done.

1:04:08

Right now, when you bring up

1:04:10

the VPN title, it

1:04:12

inserts a new default route

1:04:15

for everything and

1:04:17

points it at its virtual interface

1:04:20

so that it receives everything. What

1:04:23

it would need to do would be to send itself a

1:04:25

test ping from an IP in the user's IP

1:04:39

space and verify that its

1:04:42

virtual interface receives that

1:04:44

ping. If

1:04:46

it doesn't receive the ping, that

1:04:49

tells it something has interfered with

1:04:51

the routing between the user's

1:04:54

local host IP and its

1:04:56

own interface. So

1:05:00

yeah, that would be a cool

1:05:02

feature for a VPN to add. Meanwhile,

1:05:04

there's not really a mitigation, is there? No.

1:05:08

No. And I

1:05:10

think your use case is exactly

1:05:12

the right one, Leo, because where

1:05:14

do people deliberately bring up a

1:05:16

VPN? It's

1:05:18

when they're in a hotel, in a

1:05:20

cafe, in any untrusted environment, and they

1:05:22

don't want

1:05:27

to be sharing their traffic with everybody else. Yeah.

1:05:31

I wonder if commonly used hacking

1:05:33

tools like Wi-Fi, Pineapples and stuff are able

1:05:35

to do this. They

1:05:38

probably are. I mean, it's been around for 30 years.

1:05:41

Yeah, but well, so you mean

1:05:43

whether they're able to perform the hack. Yeah. I

1:05:47

bet that the intercepting ... That

1:05:49

seems like something you build in. Well,

1:05:51

and intercepting DHCP is such a juicy

1:05:53

target. Yeah. I mean,

1:05:55

I'll bet you that hacking tools have

1:05:57

a ... have

1:06:01

DHCP server spoofing

1:06:04

and are able to get a

1:06:06

response out immediately. Interesting. Wow,

1:06:11

this is good stuff. Thank you. Because this

1:06:13

has been everywhere, this story, and I was

1:06:15

really curious what you thought of it.

1:06:18

Yeah, so it's a problem. Again, what are

1:06:20

you going to do with the CVE? Hello?

1:06:24

Okay, I mean

1:06:26

maybe that gets it more attention. Yeah,

1:06:30

unfortunately some, apparently GPT something is

1:06:32

able to read the CVE and

1:06:34

immediately design a hack that

1:06:36

the scriptkitties could then use. So great.

1:06:41

Would you like to take a break? Is that what you're looking at

1:06:43

me like that for? I

1:06:47

know that look. We

1:06:50

will have more with Mr. Gibson in just a

1:06:52

little bit. Every

1:06:54

week there's a story or two that in

1:06:57

my mind, and I bet your mind too, you go,

1:06:59

I wonder what Steve has to say about that. That's

1:07:02

why we love you, Steve, and that's why we listen to

1:07:04

the show. We're so glad to carry the show. Today

1:07:08

our sponsor for this segment, Lookout.

1:07:11

This is kind of timely. Every company

1:07:13

today is a data company, right? We all have data

1:07:15

and we're all out in the cloud. We're all out

1:07:17

and about. Every company's at

1:07:19

risk with cyber threats,

1:07:21

breaches, leaks. This

1:07:24

is all the new normal. And

1:07:27

cyber criminals, they're getting better and more sophisticated by

1:07:29

the minute, especially with the help of

1:07:31

AI. At a time when boundaries no

1:07:34

longer exist to work, what

1:07:36

it means for your data to be secure

1:07:38

has really fundamentally changed, which

1:07:40

is why you need Lookout. From the

1:07:42

first phishing text to the final

1:07:45

data grab, Lookout stops

1:07:47

modern breaches as swiftly as they

1:07:49

unfold, whether on a device in

1:07:52

the cloud, across networks, even

1:07:54

working remotely at the local coffee shop,

1:07:57

which now we know is a little

1:07:59

hazardous. Lookout gives

1:08:01

you clear visibility into all your

1:08:03

data at rest and

1:08:06

in motion. You'll monitor, assess, and

1:08:08

protect without sacrificing productivity and employee

1:08:11

happiness for security. With

1:08:13

a single unified cloud platform,

1:08:15

Lookout simplifies and strengthens for

1:08:18

me imagining security for the

1:08:20

world that will be today.

1:08:22

Visit lookout.com today to learn

1:08:24

how to safeguard data, secure

1:08:26

hybrid work, and reduce IT

1:08:29

complexity. That's lookout.com. Thank you, Lookout,

1:08:31

for supporting Steve and

1:08:33

the work he does here. And thanks also

1:08:36

to all of our Club Twit members whose

1:08:38

donations make this show possible. You know, if

1:08:40

you find this valuable and you're not yet

1:08:43

a Club Twit member, it's just seven bucks

1:08:45

a month. We have corporate memberships too. Get your

1:08:47

whole company involved. Probably everybody in your IT department

1:08:49

should be listening to this show every darn week.

1:08:52

Go to twit.tv slash Club Twit to find

1:08:54

out more. And to those

1:08:56

of you who are already members, thank you. We

1:08:58

appreciate you supporting. On we go with

1:09:01

Mr. G. Okay, so a bit

1:09:03

of feedback. Dave Brenton tweeted,

1:09:05

Mr. Gibson, quickly may

1:09:08

I say as a machine language coder, I

1:09:10

admire your work in that area. I'm

1:09:12

a spin-right owner user and long-time fan

1:09:15

since near the beginning of security now.

1:09:17

My question is about security keys. I

1:09:19

hope this is not too long a question. And

1:09:21

it wasn't. He says I'm about to

1:09:23

make the transition to YubaKey. And

1:09:26

so I intend to purchase two to

1:09:29

have a safe fallback in case of loss.

1:09:32

I'm also planning to convert the wife

1:09:35

over to the Passkey world. My

1:09:37

question is, can the Passkeys

1:09:40

be paired across two user accounts,

1:09:43

thereby ensuring recovery in case

1:09:45

of loss with only three

1:09:47

keys. My mental model says

1:09:50

it made sense, but I do not know for

1:09:52

sure. One, can the same

1:09:54

key be applied to two different people?

1:09:57

Two, To assure full backup

1:09:59

protection? Than in all three keys

1:10:02

be coded into both users. May

1:10:05

be a silly notion but it could

1:10:07

work but could it work or should

1:10:09

I just by Four Keys to begin

1:10:11

with: Thank you for all your good

1:10:13

work and propeller had installments on to

1:10:15

Nine, Nine Nine and beyond. Yes Days

1:10:17

Zazzle I set of the beginning of

1:10:19

the show out on episode nine seven

1:10:21

three we are close it at all

1:10:23

Nine Nine we be yet we're no

1:10:25

longer fearful of that is fatal number

1:10:27

of be so sad a before you

1:10:29

get to the answer I just one

1:10:31

as well actually do the answer and

1:10:33

them. And when asked about. Missing

1:10:35

language and assembly. I have since.

1:10:39

Oh Crap. Oh okay, so I

1:10:41

chose to share Daves question. Because.

1:10:43

It's so perfectly demonstrates the

1:10:46

near total mess of the

1:10:48

user authentication world has fallen

1:10:51

into. today. It's it is

1:10:53

just a catastrophe. I'm hopeful

1:10:56

this may just be a

1:10:58

transition phase, but truth be

1:11:00

told, All. of

1:11:03

our collective experience also

1:11:05

leaves me feeling somewhat

1:11:07

skeptical. I. Worry that

1:11:09

all we have done. By.

1:11:11

Adding the Fido groups. Of

1:11:14

of I'm sorry by having

1:11:16

the Fido group lower the

1:11:18

bar for entry. From.

1:11:20

Requiring physical t dongles

1:11:23

to allowing pretty much

1:11:25

anything else Smartphones and

1:11:27

Pcs running simple software

1:11:29

pesky clients is. To.

1:11:32

Expand upon the number

1:11:34

of available options with

1:11:36

an additional and. Difficult

1:11:39

as it is to believe in this

1:11:41

day and age. Not very well thought

1:11:43

out system. A We've

1:11:46

added this new and not

1:11:48

well thought out system without

1:11:50

removing. Any of the

1:11:52

previous options. Have. Traditional

1:11:54

user name and passwords been

1:11:56

replaced. Know. are

1:11:59

they ever going to be Not in

1:12:01

this lifetime. Have the

1:12:03

I forgot my password links gone

1:12:05

away? No. Are they

1:12:07

ever going to? No. What

1:12:10

about those time-based one-time passcodes? Are

1:12:12

they going away? No. Any

1:12:14

plan for that? No. What about

1:12:17

OAuth, which brings us the log

1:12:19

on with your Google or Facebook

1:12:21

or some other account? Have those

1:12:23

been obsolete and removed? Nope.

1:12:26

Can they be? Well, not easily,

1:12:28

since many sites only know

1:12:30

their users thanks to their

1:12:33

redirection through another web services

1:12:35

authentication. And so,

1:12:37

through this pile of

1:12:39

existing half-baked remote

1:12:41

network authentication solutions,

1:12:44

we're now adding passkeys, a

1:12:47

mysterious new solution that its

1:12:49

designers all say is amazing,

1:12:52

and far more secure, which works

1:12:54

sort of like magic, right

1:12:57

up until it doesn't work at all. And

1:12:59

when that happens, what do we do? Well,

1:13:02

we fall back to send me

1:13:04

an email. What

1:13:07

we've wound up with is

1:13:09

the well-known and often observed

1:13:11

phenomenon of solution spread. We

1:13:15

invent a better idea than

1:13:17

what we had before. Perhaps

1:13:19

it's because the times have changed,

1:13:21

and the older solutions are no

1:13:23

longer adequate. Or perhaps we

1:13:26

have more technology and available processing

1:13:28

power than we had before, so

1:13:30

new solutions are available than were

1:13:32

previously. But the problem is, we

1:13:35

rarely are able to kill

1:13:37

off the things that came

1:13:39

before. Why? Because

1:13:41

by the time we can do something

1:13:43

more, too many people have

1:13:46

come to depend upon the previous solution,

1:13:49

and the one before that, and the one

1:13:51

before it. And

1:13:53

this solution spread doesn't just

1:13:55

apply to the authentication domain.

1:13:58

Just look at Windows. without

1:14:00

getting bogged down into the details,

1:14:03

every few years Microsoft comes up

1:14:05

with a new and much improved

1:14:07

way of writing applications for their

1:14:10

Windows OS. And they

1:14:12

promote the hell out of it, explaining

1:14:14

how and why it's so much better

1:14:16

than everything that came before. And

1:14:19

do they then kill off the previous

1:14:21

ways of programming Windows? No, of course

1:14:23

not. They can't. They were

1:14:26

once promoting the hell out of those

1:14:28

previous solutions and they got lots of

1:14:30

people on board using them then. So

1:14:32

even though they no longer love them

1:14:34

and are urging everyone to use the

1:14:36

new system, that never happens. I've

1:14:39

heard Paul over on Windows Weekly saying

1:14:42

that the original Windows API, Win32, should

1:14:45

have died off long ago. That's

1:14:48

what all of my Windows are written in.

1:14:51

And not just mine, a gazillion others

1:14:53

as well. And that's gazillion with a

1:14:56

g. I

1:14:58

am certain Paul knows that

1:15:00

Microsoft will never abandon Win32.

1:15:03

They can't anymore than websites

1:15:05

will ever be able to

1:15:07

stop offering username and passwords

1:15:10

with an I forgot how email

1:15:13

link. So just

1:15:15

to be clear, the industry has

1:15:17

added a bright and shiny additional

1:15:22

way for people to log

1:15:24

into their accounts. But

1:15:26

none of the existing ways are

1:15:29

or will be removed.

1:15:33

Remember that today in

1:15:35

2024, only one out of

1:15:38

every three internet users is

1:15:40

using any form of

1:15:42

password manager. I

1:15:45

really don't know what the rest are doing. These

1:15:47

are people whose iOS and Android

1:15:50

support for Passkeys is mostly aimed

1:15:52

at. These

1:15:54

people don't know, don't understand,

1:15:56

and don't care about their online

1:15:59

identity. So, when Apple or Google

1:16:01

comes along and asks, how would

1:16:03

you like to log in instantly

1:16:06

with Passkeys and never worry about

1:16:08

another password? Well, that sounds

1:16:10

great. But that's

1:16:12

not Dave. Our

1:16:14

listener whose questions launched me into first

1:16:16

taking a bit of a rant into

1:16:19

a wider view of where we stand

1:16:21

today. So, let's look at

1:16:23

Dave's situation. Dave says he's

1:16:26

planning to convert his wife over

1:16:28

to Passkeys. I'm sure

1:16:30

he means that he would like

1:16:32

to have his wife begin to

1:16:35

use Passkeys since it's not possible

1:16:37

to convert over to Passkeys

1:16:39

in any meaningful way when

1:16:42

so few websites offer the

1:16:44

option at all. The

1:16:47

caution there, since we do

1:16:49

not yet have Passkey transportability,

1:16:52

is to be careful about which

1:16:54

app is holding a site's Passkeys.

1:16:56

As I mentioned last week, iOS,

1:16:59

Windows, Android, and now an

1:17:01

increasing number of traditional password

1:17:04

managers will all be vying

1:17:06

to be the app

1:17:09

that generates the Passkey to

1:17:12

be provided to a website. Since

1:17:15

only that app will

1:17:17

then be able to authenticate the user

1:17:19

to that site with a Passkey, the

1:17:22

only sound strategy will

1:17:24

be to only and

1:17:26

always use a single

1:17:28

platform for Passkeys. This

1:17:33

issue and Dave's other questions

1:17:35

require a quick bit of

1:17:38

foundation about the operation of

1:17:40

Passkeys. When

1:17:43

an application prompts its user

1:17:45

about whether the user wishes to

1:17:48

have it create a Passkey, that's

1:17:51

exactly what's happening. The

1:17:53

application generates a

1:17:55

cryptographically strong secret and

1:17:58

private key which

1:18:00

never leaves the application

1:18:03

and which the application guards carefully.

1:18:05

And in this case, I'm using

1:18:08

key and application interchangeably. From

1:18:10

that closely held private key,

1:18:13

it then generates a public key.

1:18:16

And only the public key is

1:18:18

sent to and retained by the

1:18:21

website. In the

1:18:23

future, that website will use

1:18:25

the public key it

1:18:27

holds to verify the signature

1:18:29

of a challenge that it sends to

1:18:31

the user's past key authenticator. So

1:18:34

my point here is that

1:18:36

today there is no

1:18:38

provision for these private

1:18:40

keys which were

1:18:43

generated internally and have

1:18:45

ever since been guarded by the

1:18:47

application to ever leave

1:18:49

that application's control. And

1:18:51

a security conscious organization like

1:18:54

Apple can make the

1:18:56

defensible claim that since

1:18:58

all of the past key's security

1:19:01

derives from the secretness of

1:19:04

these private keys, which

1:19:06

is crucial, no

1:19:09

other application, including its

1:19:11

user, can or

1:19:13

should be entrusted with

1:19:15

their stewardship, with the stewardship

1:19:17

of the past key's private

1:19:20

key. Since this

1:19:22

represents a powerful platform

1:19:24

lock-in, it's not at all

1:19:26

clear to me that Apple will ever allow

1:19:28

for past key's export. That

1:19:30

being the case, I

1:19:33

think that a very strong case can be

1:19:35

made for only ever

1:19:38

storing past keys in a third party

1:19:40

past key's client, such as a browser

1:19:42

extension. In theory, it ought to

1:19:44

be possible for a website to allow

1:19:47

its user to replace one past

1:19:49

key with another, so if

1:19:51

Apple or Android were to inadvertently become

1:19:53

the generator and holder of a past

1:19:56

key, if a

1:19:58

website supported past key replacement,

1:20:01

it should be possible to migrate

1:20:03

away from one passkey application to

1:20:05

another. And if, and I was thinking

1:20:07

about this, if a

1:20:09

website doesn't explicitly allow you to

1:20:12

migrate between passkeys, hopefully it allows

1:20:14

you to delete a passkey, in

1:20:16

which case your account would not

1:20:18

be associated with one, and then

1:20:20

you could re-associate it with a

1:20:22

passkey from the provider that you're

1:20:24

wanting to switch over to. So,

1:20:29

so the, the,

1:20:32

the real point here is that it

1:20:36

is the application that generates the

1:20:38

passkey. It is never something that

1:20:40

we're able to supply from the

1:20:42

outside. So, just

1:20:45

to put a bit of frosting

1:20:47

on this discussion before we talk

1:20:49

about the platforms with hardware, with

1:20:51

hardware authentication doggles, I

1:20:53

wanted to share a few points from

1:20:56

Google's Chrome FAQ.

1:20:59

This is Google's Chrome browser

1:21:01

FAQ about passkeys. They

1:21:04

start off, of course, with all the

1:21:06

glowing bits. Under manage

1:21:08

passkeys in Chrome, they say

1:21:11

you can use a passkey to

1:21:13

sign in easily and securely with

1:21:16

just a fingerprint, face scan, or

1:21:18

screen lock. Passkeys

1:21:20

are a simple and secure way to sign in

1:21:22

to both your Google account

1:21:24

and all these sites and apps you

1:21:27

care about without a password. You

1:21:29

may be asked to sign in to

1:21:32

a website with a passkey or create

1:21:34

one to improve your account's security. Then

1:21:36

they have a little tip. Passkeys are

1:21:39

built on industry standards, so

1:21:41

you can use them across

1:21:43

many platforms. I love those

1:21:45

industry standards. Oh,

1:21:47

Leo, that's the happy news. That

1:21:50

all sounds terrific. And of course,

1:21:52

we ask, what here? What could

1:21:54

possibly go wrong? Well, here's what

1:21:56

Google had to say about that.

1:22:00

under store pass keys in

1:22:02

Windows. They said if

1:22:04

you have Windows 10 or up you

1:22:06

can use pass keys. To

1:22:09

store pass keys you must set up

1:22:11

Windows Hello. Windows Hello

1:22:13

does not currently support synchronization

1:22:15

or backup. So pass keys

1:22:17

are only saved to your

1:22:19

computer. If your computer is

1:22:21

lost or the operating system is reinstalled,

1:22:24

you cannot recover your pass keys. Oops.

1:22:28

Or store pass keys in Mac

1:22:30

OS. You can save pass keys

1:22:32

in your Chrome profile where

1:22:35

they're protected by a Mac

1:22:37

OS keychain. Then under important they

1:22:39

said Chrome cannot save or

1:22:41

use pass keys stored in iCloud

1:22:44

keychain. If your computer is

1:22:46

lost or your Chrome profile is

1:22:48

deleted, you cannot recover your

1:22:50

pass keys. And third,

1:22:53

you can use a security key to

1:22:55

store your pass keys. Important, pass

1:22:57

keys stored on security keys are

1:22:59

not backed up. If you lose

1:23:02

or reset the security key, you

1:23:04

cannot recover your pass keys. What

1:23:07

a wonderful system. This clearly

1:23:09

represents a huge leap

1:23:12

forward. Psi. Wow.

1:23:15

It's clear that unfortunately what

1:23:17

we have at the moment

1:23:19

is an extremely fragile system.

1:23:23

The problem is the extreme

1:23:25

secrecy surrounding the private

1:23:27

keys which create the

1:23:29

pass keys. It's true that they

1:23:31

do not... I'm sorry.

1:23:34

I'm sorry. It's true that they do need to

1:23:36

be guarded. Unfortunately at the

1:23:38

moment they're being jealously guarded. How

1:23:42

Microsoft could possibly

1:23:44

imagine that it's practical

1:23:46

to have all of a user's

1:23:48

pass keys locked up in a

1:23:51

single machine unable to

1:23:53

synchronize with any of a

1:23:55

user's other's devices is beyond

1:23:57

me. But we're

1:23:59

ready. Ready to entertain the second part of Dave's

1:24:02

question where he asked, can the

1:24:04

past keys be paired across two user

1:24:06

accounts, thereby ensuring recovery in case of

1:24:08

loss with only three keys? He

1:24:11

says, my mental model said it made sense, but I

1:24:13

do not know for sure. Can

1:24:15

the same key be applied to two different

1:24:17

people to assure full backup protection? Can all

1:24:20

three keys be coded into both users? The

1:24:24

answer is that

1:24:26

not one of

1:24:28

those operations Dave

1:24:31

is asking for is

1:24:33

available, not one. And

1:24:36

what's more, I just double

1:24:38

checked. As we learned last

1:24:40

week, YubaCo's YubaKeys have the

1:24:42

most ample storage for past

1:24:45

keys of any hardware

1:24:47

past key dongle in the

1:24:50

industry. And even it

1:24:53

is limited to a total of only 25.

1:24:57

And they are utterly and absolutely

1:25:00

non-exportable. A

1:25:03

YubaKey is at its heart

1:25:05

an HSM, a hardware security

1:25:07

module. The internal

1:25:09

YubaKey dongle hardware contains

1:25:12

a very high entropy

1:25:14

random number generator that's

1:25:16

used to synthesize a unique private

1:25:18

key. That

1:25:21

private key never leaves the

1:25:23

device. There is no way

1:25:25

to export it. The exportation

1:25:27

does not exist. There's

1:25:31

no way to put a past key

1:25:33

in and no way to take a past key

1:25:35

out. This would not be

1:25:37

a problem if sites were to allow multiple

1:25:39

past keys to be registered

1:25:41

for a single account. And

1:25:44

there's no reason that would not be possible,

1:25:47

but how many sites today

1:25:49

support the use and management

1:25:51

of multiple passwords for a

1:25:54

single account? I've

1:25:56

never seen one. So it's

1:25:58

unclear why so many people are using a past key. Support for

1:26:00

multiple pass keys would ever

1:26:03

be created even though nothing

1:26:05

prevents it. With

1:26:07

Yuba keys having a 25 pass

1:26:09

key limit other than

1:26:12

for experimentation, they seem

1:26:14

most practical for higher-end

1:26:17

enterprise-grade security applications and

1:26:20

perhaps for eventually signing into

1:26:22

only a few of the

1:26:25

most secure sites where the

1:26:27

inconvenience of having an absolute

1:26:29

hardware lock is warranted by

1:26:32

its ultimate level of hardware-level

1:26:34

security. And as we noted

1:26:36

last week, a Yuba key might be used

1:26:38

to unlock a password manager, which is

1:26:41

where we would all have

1:26:43

to conclude all of a

1:26:45

user's pass keys should probably be stored.

1:26:48

The only sane conclusion we can

1:26:50

draw is

1:26:53

that while this is all

1:26:55

very interesting, none of this is

1:26:57

yet ready for prime time. Poke

1:27:00

at it, experiment with

1:27:02

it, but wait until

1:27:04

Bitwarden's pass key supporting

1:27:06

mobile clients emerge from

1:27:08

their current beta testing

1:27:10

state, at which point

1:27:12

it will be practical

1:27:14

to start depending upon

1:27:16

pass keys because they

1:27:18

will be in a

1:27:20

single sane multi-platform client.

1:27:23

And Bitwarden, which is, we should say, a

1:27:26

sponsor of the Twit network, will

1:27:28

likely be offering backup

1:27:31

and support and

1:27:34

exportation of those once

1:27:36

the security protocol for doing

1:27:38

that, which is reportedly underway

1:27:41

within the FIDO group, is

1:27:44

concluded. So Bitwarden

1:27:47

will then generate and

1:27:49

hold our pass keys even when

1:27:51

other pass key clients on iOS or

1:27:54

Android might be trying to. And then

1:27:56

of course, as we said last week,

1:27:58

the challenge is... making sure

1:28:01

that your chosen

1:28:04

PASCE authenticator is universally used

1:28:06

even in an environment where

1:28:08

multiple authenticators are all vying

1:28:11

for attention. I

1:28:14

have to say, the reasonable

1:28:17

things that people would want to

1:28:19

do are not available.

1:28:22

They cannot be done. Wow. Yeah.

1:28:26

Yeah. I

1:28:29

saw there was a Hacker News story about,

1:28:31

somebody wrote about why it's 100 times harder

1:28:34

to implement PASCEs on your website than you might

1:28:36

imagine. I think this is going to be,

1:28:38

I feel like people

1:28:41

are going to throw up their hands and say, okay,

1:28:43

fine, never mind. And

1:28:45

that's depressing. Right. Right.

1:28:48

And as we said last week, if it doesn't

1:28:50

achieve critical math, then it'll just

1:28:53

be, exactly as one

1:28:55

of our listeners said, or no,

1:28:57

no, it was the

1:28:59

guy who did the Rust

1:29:01

web-authent client. He said, I

1:29:04

feel that this will, you know, it'll

1:29:06

be like ad blockers. A small percentage

1:29:08

of people take the trouble to do

1:29:10

it, but it's sort of a niche

1:29:12

and it never really becomes a problem

1:29:14

for ad companies. And in this

1:29:16

case, it just never takes hold.

1:29:20

So speaking of- It is a mess. It is

1:29:22

a mess. And it's not getting any better.

1:29:24

Rust did not solve it. We've been

1:29:26

trying, I mean, I remember when Microsoft tried the

1:29:28

single sign-on thing 20 years

1:29:30

ago. We've been trying to solve it. And

1:29:32

they had something called Passport. That's what I was talking about.

1:29:35

Passport, exactly. It was a single sign-on. And

1:29:37

it didn't get adopted and that's that. And-

1:29:41

Yep. Oh, well. Oh,

1:29:43

well. So one last

1:29:45

piece of feedback from Will you stop? Before you

1:29:47

do that, can I answer you a question? Yeah.

1:29:49

About assembly language. Yeah. I was talking

1:29:51

the other day about how one debugs

1:29:53

in a higher level language. You'll write a print

1:29:56

statement, for instance, and it'll tell you

1:29:58

all your stuff. You must have some- macros you've

1:30:00

written over the years to help you debug

1:30:02

assembly or do you? No.

1:30:07

I knew it. You

1:30:11

just write it right the first time. Well,

1:30:14

so for not for debugging,

1:30:16

but for example, one

1:30:20

of the reasons it would be

1:30:22

difficult for me to share my assembler

1:30:24

is that I have built up a

1:30:27

macro archive of things I do.

1:30:30

For example, I use a macro

1:30:32

called zero and

1:30:34

it takes a

1:30:37

register name. Well,

1:30:39

it simply expands to XOR register comma

1:30:41

register. Right, to zero it out. Because

1:30:44

you know when you XOR something, exactly,

1:30:46

when you XOR something with a self

1:30:48

you get zeros and it's very fast

1:30:51

because it doesn't depend upon a memory

1:30:53

fetch or the previous data

1:30:56

or the previous contents of the

1:30:58

register. The

1:31:00

point is if I wrote XOR

1:31:03

something comma something, I would

1:31:06

have to look at it and say

1:31:08

okay XOR and then look at what

1:31:10

am I doing and then realize oh,

1:31:13

I'm wanting to zero that. Well, it's much

1:31:15

better if I just say zero and then

1:31:17

the thing. Anyway,

1:31:21

and you cannot do that for

1:31:24

variables. That is the

1:31:26

Intel architecture will not allow you

1:31:28

to XOR memory

1:31:30

with another memory. You can only

1:31:33

XOR register with a register or register

1:31:35

with memory, but not memory with memory.

1:31:38

I have a variable, I use the

1:31:40

macro reset which moves

1:31:43

a zero into it. But you don't

1:31:45

have any macros for kind of displaying

1:31:47

the contents of the stack purely

1:31:50

for debugging. You don't have anything like that. You just

1:31:52

look at the code and figure out what's going on.

1:31:55

Oh, no, no. So, I definitely have a

1:31:57

debugger. Oh, God. Oh, yeah, yeah.

1:32:00

Yeah, yeah. So... Masm comes with

1:32:02

a debugger, right? Or no? So,

1:32:04

Masm doesn't, but there were back in

1:32:06

the day a bunch of third party

1:32:08

debuggers. I use something called Periscope, which

1:32:11

was written by a guy named Brett Salter

1:32:13

years ago. I don't remember that yet. He

1:32:15

passed away a few years ago. There was

1:32:17

also something called Softus, which

1:32:20

was an ice stands for

1:32:22

in-circuit emulator. And in

1:32:25

the really old days, you would

1:32:27

pull the processor off the motherboard

1:32:29

and plug in this paddle that

1:32:32

then had a cable running through a

1:32:34

bunch of things that emulated the processor

1:32:36

that allowed you essentially to get inside

1:32:39

the processor. Wow. That's

1:32:41

wild. So that was called an ice,

1:32:43

an in-circuit emulator. And so Softice was

1:32:45

essentially using protected mode to

1:32:47

do all the same sorts of

1:32:49

things. So there have absolutely always

1:32:51

been debuggers. And one of the,

1:32:54

one of the banes of

1:32:56

developing for Spinrite was that

1:32:58

I'm, you know,

1:33:00

DOS and 16 bits. And

1:33:04

it was very difficult to create

1:33:06

an environment where I was

1:33:08

able to have networking in order for

1:33:11

my code to get down into the

1:33:13

target machine and debugging at the same

1:33:15

time. So one of

1:33:17

the things I'm really looking forward to as I move

1:33:20

to my own environment is,

1:33:23

for example, this RTOS32

1:33:25

that will be the home for

1:33:28

Spinrite 7, it works

1:33:30

with Visual Studio transparently. So

1:33:33

I get to just live in

1:33:35

a really nice, gooey IDE and

1:33:37

do all of my debugging. And

1:33:40

what's really cool, Leo, I bought so

1:33:42

many motherboards and so many random hard

1:33:44

drives through eBay when our

1:33:47

testers were reporting that on my

1:33:49

Gymcrack 27Z, it does such and

1:33:53

such. And I was like, oh

1:33:55

my God. So I'd have to go look, I'd go

1:33:57

to eBay, search for Gymcrack 27Z. and

1:34:04

I would buy one and so my amazing

1:34:07

wife put up with having motherboards everywhere.

1:34:09

Oh, for the dining room table, I'm

1:34:11

sure. So what's very

1:34:13

cool about RTOS32 is

1:34:16

it allows internet, trans-internet

1:34:18

debugging. Oh, nice. So

1:34:21

if something is happening

1:34:23

on that guy's JimKrak27Z,

1:34:26

I'll be able to actually have him

1:34:28

contact me and debug it on

1:34:31

his machine. Oh, that's really cool.

1:34:34

Wow. Yeah. Very

1:34:36

neat. Oh, okay. So you have some pretty

1:34:38

good tools, it sounds like. Oh, yeah. And in

1:34:40

fact, one of the

1:34:43

things that I've learned is invest

1:34:45

in your tooling infrastructure before

1:34:47

you do anything. It

1:34:49

is so nice to have a

1:34:52

convenient debugging environment. Absolutely.

1:34:56

On we go. I'm sorry, I didn't mean to

1:34:58

interrupt. I was just curious. I was debugging the other night

1:35:00

and I was thinking, I wonder how Steve does this. Now

1:35:02

I know. Yeah, you absolutely have to

1:35:05

have a good deal. It allows you

1:35:07

to see the stack and the contents of the

1:35:09

registers and what's in

1:35:11

memory and what your local

1:35:13

variables are. All of that is

1:35:15

made really very nice with Visual

1:35:18

Studio. Nice. Okay.

1:35:20

Willie Scott. He says, okay,

1:35:23

he has some feedback and advice

1:35:26

about the operation of the iCloud

1:35:28

Keychain and I bet you

1:35:31

he knows what's going on or at least

1:35:33

gave us enough of a clue. He said,

1:35:35

hi, Steve. In regards

1:35:37

to your discussion of pass keys

1:35:40

on last week's show, the part

1:35:42

about the author's partner losing her

1:35:45

iCloud Keychain passwords intrigued

1:35:48

me. After the

1:35:50

last pass hack, I decided to

1:35:53

switch to using iCloud Keychain for

1:35:55

my passwords because I'm in the

1:35:57

Apple ecosystem and wanted to start

1:35:59

using using pass keys instead

1:36:01

of passwords wherever possible. I'm

1:36:04

writing to mention that I

1:36:06

too have had

1:36:08

passwords and two-factor authentication

1:36:10

codes wiped from my

1:36:13

iCloud keychain. Although

1:36:18

my keychain has never been

1:36:20

fully wiped like the poor

1:36:22

partner's keychain did. As

1:36:24

near as I can tell, I believe I

1:36:26

know the culprit of why it may

1:36:29

be wiping credentials from iCloud keychain and

1:36:31

wanted to pass this along to anyone

1:36:33

who might still be using iCloud keychain

1:36:35

to store their passwords or

1:36:37

who knows somebody who may. When

1:36:41

I started changing all my passwords and

1:36:43

adding accounts into iCloud keychain, I

1:36:46

noticed that an old

1:36:48

Amazon password that I don't

1:36:50

use anymore was already stored

1:36:52

in there, probably

1:36:55

from when the Amazon app asked, Do

1:36:57

you want me to remember your password?

1:37:00

It was an old password that I don't use

1:37:02

anymore so I deleted it. However,

1:37:04

a couple of days later, I noticed

1:37:07

that even though I deleted that

1:37:09

password, or so I thought, it

1:37:12

had somehow reappeared in

1:37:14

my iCloud keychain. Not

1:37:16

only that, but I also noticed that

1:37:19

one or two accounts that I had

1:37:21

recently added to the keychain were missing

1:37:24

and this process repeated itself a

1:37:26

few more times. So

1:37:29

that's when I started investigating. While

1:37:31

digging through the settings, I

1:37:33

went through my Apple ID account

1:37:35

settings and that's when I realized

1:37:38

that my old iPhone

1:37:41

6S Plus, which

1:37:44

was running an old version of

1:37:46

iOS, iOS 14

1:37:48

to be exact, was still

1:37:50

signed into my iCloud account

1:37:53

and had iCloud keychain turned

1:37:55

on. I

1:37:57

removed that old iPhone from my iCloud account.

1:38:00

out account and ever since I

1:38:02

did that no passwords

1:38:04

have been wiped. If

1:38:07

you are in an Apple ecosystem it

1:38:09

is always a good idea to keep

1:38:11

your devices up to date but it

1:38:13

might also be a good idea to

1:38:15

do some spring cleaning and remove old

1:38:17

Apple devices from your iCloud that you

1:38:19

don't use anymore. Having

1:38:22

said all that I sadly was

1:38:24

agreeing with a lot of the points

1:38:26

you were making about Passkeys

1:38:29

and I think I've decided

1:38:31

that I will probably switch over

1:38:33

to Bitwarden once Passkeys become officially

1:38:36

supported in Bitwarden using

1:38:38

and he says https://bitwarden.com. Thank

1:38:42

you. Of course. Yes

1:38:44

sir. Special sponsor link. Which

1:38:47

I think we are about to talk about. Yes we are actually.

1:38:50

Thank you for a great show. I look

1:38:52

forward to it each week. I'm also a

1:38:54

proud Spinrite owner and can't wait to start

1:38:56

using 6.1 on my SSDs and a troubled

1:38:58

hard drive. So this

1:39:02

mysterious iCloud credential

1:39:04

removal has all

1:39:07

the feel of

1:39:09

something Apple would be

1:39:11

deliberately doing out of

1:39:13

their typical abundance of

1:39:15

caution. I'll bet there's

1:39:18

a security model behind it. For

1:39:21

example while an older

1:39:23

iPhone is also

1:39:25

signed into an accounts

1:39:28

iCloud keychain Apple might

1:39:30

be deliberately limiting what

1:39:32

they're willing to save

1:39:34

into that shared keychain

1:39:36

while an older and

1:39:39

presumably lower security device

1:39:41

also shares access. In

1:39:43

other words it's a feature not

1:39:46

a bug. I

1:39:50

guess it could be that. I

1:39:52

don't like that kind of unexplained behavior

1:39:54

however. It sounds like

1:39:57

Apple though to say oh we're

1:39:59

not going to let you hurt yourself, you

1:40:01

know, we're gonna delete your, you know,

1:40:05

the keys you've just saved because

1:40:07

otherwise one of your insecure devices

1:40:09

might get them. Ay yay yay.

1:40:12

Yeah, I'll make sure that you always should

1:40:14

remove old devices. That's maybe why I've never

1:40:16

run into this. I always remove the old

1:40:18

devices. So hmm. Yep.

1:40:21

Very interesting. I do happen to

1:40:23

have an iPhone 6 right here.

1:40:25

Wow, look at that. That doesn't

1:40:28

work anymore. Look at that

1:40:30

home button and think fondly on it

1:40:32

because Apple has, as of today, discontinued

1:40:34

all the devices that had

1:40:36

home buttons. The last one

1:40:38

you could buy was the iPad base model

1:40:41

and that's now been superseded. So the

1:40:43

home button is officially a thing of

1:40:45

the past, as is the

1:40:48

headphone jack, I think. I

1:40:50

think with it all, facial recognition?

1:40:52

Yeah. It's all Face

1:40:54

ID now. Yep, makes sense.

1:40:58

Let's talk about it, Warden, and then I

1:41:00

want to talk about your subject matter for

1:41:02

the day. Indeed. What's

1:41:05

going on with Google in the UK? What

1:41:07

is happening here? Why are

1:41:09

they putting off the third party cookie?

1:41:12

Not so fast. What's the deal, man?

1:41:15

This portion of the show brought to you by Bit Warden.

1:41:17

The password manager I use, the password manager

1:41:19

Steve uses, and pretty much everybody I know

1:41:21

uses. We were talking Mac Break Weekly. There's

1:41:24

something that happens in the geek community where we, without

1:41:27

coordinating with one another,

1:41:29

converge on the best solution.

1:41:32

Neelai Patel, the laser printer that

1:41:35

everybody uses but nobody mentions, the

1:41:38

screen cleaner we all use but nobody mentions.

1:41:41

I would say Bit Warden has

1:41:43

become like that. The password manager

1:41:45

of choice for anybody who's

1:41:47

really paying attention. Now,

1:41:49

there's a good reason for this. It's open

1:41:51

source. It's

1:41:53

feature rich. It is

1:41:56

great for home, great for office. They have

1:41:58

a Teams and an enterprise. account.

1:42:02

They support Yuba keys and now good

1:42:04

news they just announced they officially support

1:42:07

Pascis. They've supported it for a while but now

1:42:09

they support on the browser extensions and

1:42:11

mobile devices. So it's a really

1:42:13

good solution if you're saying I don't want

1:42:15

to have my Pascis tied to a physical

1:42:17

device use Bitwarden for your

1:42:19

Pascis. I've been doing this for a while and now

1:42:22

everywhere you go on Android

1:42:25

iOS Mac Windows Linux you've got

1:42:27

your Pascis. Pascis

1:42:29

on mobile are available on iOS open beta

1:42:31

going on right now on Android. We'll talk

1:42:33