Yeah, so I wanted to just try to give some practical advice.
It's applicable really to any of you,
regardless of your expertise level.
And one of my many, many random things
that I've researched over the years
is this topic of backing up your seed phrase.
While my day-to-day job is helping people
keep their assets and their private keys safe,
backups are an important part of this
because it's really about disaster recovery.
So why are we backing up seed phrases?
Well, hopefully, if you have a nontrivial amount
of crypto assets, you're using some sort of hardware device
like a Trezor ledger.
Take your pick to keep those keys off the internet.
That basically protects you from like 95%
of the things that could go wrong with people
trying to steal your money and get your keys.
But what happens if that device gets lost, stolen, blows up,
melts down?
These things are kind of fragile.
Stuff can happen.
You need to have a way to go get a new device
and load your keys into your wallet.
And while it may make sense that when you unpack
one of these devices, it gives you this little piece of paper.
And it says, hey, write down your seed phrase.
I actually find it kind of concerning.
I think a lot of these wallet manufacturers
have kind of cheaped out.
Because if you think about it, what
is a piece of paper really going to protect you against?
The only thing that that piece of paper does well
is keep those keys off of the internet.
You definitely don't want to be taking a picture of your seed
phrase and putting it into your cloud storage.
A lot of people have lost their money that way.
You don't want to back up your seed phrase
anywhere on a computer or internet-based device.
But if you're keeping it offline,
there are other things that can go wrong.
So one of the things that I've come up with
is a sort of threat model for some of the things
that can go wrong.
And this is basically physical destruction.
So house fires are actually more common
than I think a lot of people realize.
Corrosion can happen in environments.
For example, if you have storms or some sort of water
or drainage leak.
And crushing, of course, can happen
if wherever you're storing that backup happens to collapse.
So I have a number of tests that I
have done on about 80 different backup devices
over the years where I've tried to figure out exactly what
are the best attributes of these backups.
And like I said, there's probably
over 100 different retail level backup devices on the market.
I've tested about 80 of them.
And you can really categorize them
in just a handful of different ways.
You've got about half a dozen different form factors.
So you've got your standard plate.
It's just a single square or rectangular piece of metal,
usually stainless steel.
Then you've got your tile-based backups
where they give you these pre-engraved or pre-stamped tiles
that you just sort of slide into a cassette-like device.
Then you've got other sort of rod-shaped, very strong
and sturdy pieces of metal that you can engrave or stamp.
There's a common form of do-it-yourself
where if you're worried about privacy aspects
and not buying something that you
have to get shipped to you and then potentially have privacy
leaks, like has happened with a number of retailers
in this space, there are ways you can just go to the hardware
store and buy some washers and a bolt and screws
and basically make your own pretty decent backup
for just a couple bucks.
And then some other variations where you might even
have rods with tiles in them and put those into a pipe
and just try to make the thing more robust or stack disks
on top of each other.
But then regardless of what form factor you have,
there is a number of different ways
that you can actually inscribe this data.
And the different inscription types
also has different attributes of what they can be robust against.
So the common ones that you'd normally see
are stamping where you literally have this rod with a letter
and you have to take a hammer to it.
And if you've never done that before,
I highly recommend not ever doing it
because it's a terrible, terrible process.
You'll probably end up injuring yourself
if you're not really careful.
And if you're like me, who has tested dozens and dozens
of these devices, if you don't wear ear plugs,
then after doing a couple thousand strikes,
you're going to have some inner ear pain.
Center Punch is one of my favorite methods.
You can kind of see them at the bottom down there.
It's literally, it looks like a pin and all you have to do
is just kind of put a little pressure on it
and it puts a divot into whatever piece of metal you're on.
There's a couple different ways of doing etching,
either by hand with a really sharp pin type device
or with a Dremel, which of course is electric engraving
and can get a bit more force on it.
Or you can get pre-engraved tiles, of course,
which are very easy.
And so just sort of from a perspective of usability,
if you're out there surveying the market
and trying to figure out what's going
to be more user friendly, this is kind of the list
of what I find easier, but as we'll get into,
these are not necessarily congruent
with the robustness aspects of it.
So stamping is probably the most common,
but like I said, I really hate it.
I don't recommend it.
It is very robust, but it's not easy to do right.
So what do I do when I test these devices?
Like I said, I've got my Threat model, we've got heat.
I have this propane torch that I've measured
that gets over 2,000 degrees Fahrenheit.
Your average house fire temperature
is between 1,500 and 1,700 Fahrenheit.
So I throw that in the flame for 20, 30 minutes.
Make sure it stands up well.
Acid, I don't have months and months to spend
of doing a salt water test.
So I literally use hydrochloric acid
to sort of accelerate that.
I figure if it can withstand a really, really strong acid,
it'll be able to withstand something like Brian water.
And then I've got a 20 ton hydraulic press.
I figure if your house collapses,
probably not gonna get too much more weight on it
than that.
Now maybe if you're in the very bottom
of a high rise building or something.
So if you really wanna see all of the details
of this stuff, you can check out this URL
or scan this QR code to go there.
And this is like all of my 80 devices.
I have them all ranked by each of those three categories,
sort of A through F grades.
And at the very top, it'll basically have the best
like 20 or so devices that really have straight A grade
ratings for everything that I have tested.
So what are the most common failures that we see?
Well, what I found is a number of these devices
could not even withstand a house fire temperature.
Several of them cheaped out
and they actually are made out of aluminum
instead of stainless steel.
Aluminum has a melting point only around like 1,200 degrees.
So obviously that ended up in a pile of slag.
Another thing that I've noted is that the devices
that are more complicated, that have more moving parts,
especially the ones that have tiles
and things that you slide in and out of them,
they tend to either have fasteners that fail
or they will just warp and deform
as they're heated and cooled off.
Because as part of my heating test,
I also dunk the red hot device into a bucket of water
to simulate fireman coming through with a fire hose.
You need to have both the extreme heat
and then the extreme cool off
and make sure the whole thing won't just blow up on you.
Another issue is that even if the device itself survives,
you have to worry about, okay,
what is the template on the device set up as?
So I've seen some devices where instead
of engraving the template, they just print it on with ink.
Well, as you can see here, we've got a device,
it's got data inscribed on it, the data is still there,
but good luck reading that
if you can't go and find the original template.
And then finally, we've got some issues
with some of the devices where the device itself stays intact,
but due to the way that it is actually created,
if it's deformed in the right way,
some of these tube-based devices,
you literally can't open them up again.
So you would have to go find some really heavy-duty
shop equipment in order to be able
to retrieve the data out of there.
So what are the strongest backups in my opinion?
You want stainless steel or titanium,
you really want it to be a small
but thick solid piece of metal,
and I highly recommend using that center punch
with the divots.
Electrolysis is also really good,
but that's not something that you come across very often.
And then of course, you wanna make sure
that that template is similarly inscribed
or etched into the device.
Now, this is all good and well,
but I ran into a problem about a year ago
where I had a seed phrase backup that I needed to destroy.
And so I had some metal plates,
and I had both stamped and Dremel etched
the seed phrase into them.
And the reason I needed to destroy it
was that I wanted to make it more robust,
and I basically wanted to use a process called Seedsor
to split that backup so that it no longer existed
only in one location.
Instead, it's now, it's a two-of-two type of backup.
So we have multiple copies of it geographically distributed,
and that's because I wanted to improve
the sort of physical attacker security characteristics
against it.
That's a whole other issue.
We won't even really have time to get into,
is that if you do have a really strong backup,
if somebody comes across it
and you don't have any sort of encryption or sharding,
then that's problematic because that attacker
is probably gonna be able to read the seed phrase,
load it up into a wallet and steal all of your money.
So how do you destroy one of these backups?
Well, if you're willing to forego
some of the robustness characteristics,
just use one of the tile-based backups.
You open that up, you dump the tiles out,
and you scatter to the wind, and you're done.
But as I've seen from my tests,
the tile-based devices are a lot more likely
to also get destroyed and scattered
during a catastrophic sort of experience.
So what if you do have one of these robust,
solid pieces of metal?
That's the problem that I ran into.
And essentially, I started whacking away at it.
I said, you know, I can't destroy the metal itself,
but what I need to do is I need to make it illegible.
I need to make whatever the data is
that's inscribed onto this thing,
such that if an adversary gets ahold of it,
they'll have no idea what's on there.
So what I really found out very quickly
is that with stamping, you'd basically have to spend
at least 10 times as long trying to destroy
this device as you did setting it up,
because you need to go and you need to grab
a bunch of different stamps,
and you need to hit every single one of those letters,
and just put so many different letters onto it
that it's impossible for an attacker
to be able to discern what the original letters were.
Etching is also really hard.
I spent probably a couple of hours with the Dremel.
Once again, thankfully, with earplugs in,
this was a terrible experience,
trying to overwrite and overwrite,
and if you know much about forensic analysis
and data recovery, then you may understand stuff of like,
for example, how difficult it is
to truly scratch a serial number off of a firearm.
If you scratch a serial number off of a firearm
and law enforcement gets ahold of it,
they can take it to a lab and they can do
sort of microscopic level analysis
and still be able to see really deeply
what those original numbers and letters were.
And then eventually it kind of hit me, center punch.
It's so easy to do because if you want to destroy
or make illegible one of these center punch plates,
all you have to do, it takes a couple of minutes.
You just put divots in all of the other squares on that grid,
and now there's no way of discerning
which of the divots was original
and which one was added later.
So I would say the undisputed winner,
both in terms of robustness for these metal backups
and for the optionality to be able to destroy them
or get rid of them or make them obsolete and deprecated,
has to be just a single steel plate with center punch
because even though the device itself is robust
and can't be physically destroyed,
you easily have the option to make that illegible if you want.
So, questions.
Everybody's mind has been blown.
Great, yes.
So the plate you're talking about, both the punch one,
what is it made of?
So there are probably four or five different plates
that are like that on the market.
There's like seed plate and block plate,
and these are all stainless steel,
generally four to five millimeters thick.
I would recommend at least four millimeters thick
because if it's less thickness than that,
then it's a lot easier to deform in the crush test.
And then if you really wanna get deep into the weeds,
there's of course many different types of stainless steel.
If you can find one that is a marine grade stainless steel,
then that will be even better with the corrosion resistance.
And then what, how much effort is needed to punch that punch?
Do they make them out of titanium or just...
Yeah, so the titanium plates.
Tungsten, I haven't seen anyone do tungsten.
Titanium, you can also do center punch.
But really, it probably is only like three to four
pounds of pressure to do that center punch.
So even grandma can do it.
Maybe on top of the way to keep them.
Yeah.
And if you explored like Bip.89 or Bip.89,
where you do the separation on multiple plates.
Yeah, yeah, so that's one of the things
that we haven't gotten into.
So you can, for example, there,
originally there was like Bip.38 encryption.
There's also Bip.39 passphrase,
where there are ways to like add additional data,
and then you can have that additional decryption data
basically stored somewhere else.
Or you can do what you were saying,
is it, forget if it's Bip.85 or Bip.89,
but there are a number of different protocols out there
for distributing these backups.
And where do you put them is, of course, the real question.
Personally, I recommend a distributed type of setup
between different places
that you have access to,
places that trusted family members may have secure.
And if you're doing it right,
even a bank vault will work.
The downside to if you're putting it somewhere
like in a safety deposit box,
which by the way are becoming a lot rarer these days,
you only wanna do that or put it
in a like professionally managed safety deposit box
if you have it split or encrypted.
Because you cannot be 100% sure
that that safety deposit box won't be opened
by an employee or even law enforcement or whatever.
There have been a number of cases
where things have gone wrong there.
But the main thing is you're distributing these geographically
because the whole point of the backups
is to be able to recover from a disaster
that destroys your main wallet
and your main devices where you're keeping
those seed phrase active for regular use.
All right, well, I'm glad to know that no one here
is ever gonna lose any of their crypto.
This is what I've spent the past 10 years
trying to help people out with.
Thank you.