Latest / The FIT4Privacy Podcast | Exploring ISO Standards AI Privacy | Grow Skills Store / Is AI Creating New Privacy Challenges E161

Transcript

Is AI Creating New Privacy Challenges E161 · The FIT4Privacy Podcast | Exploring ISO Standards AI Privacy | Grow Skills Store

• 0:00
  When you talk about privacy challenges, we more often hear
• 0:04
  that the risk of the challenge is accelerating because of
• 0:09
  artificial intelligence. Is that the case you believe in
• 0:12
  also? Yeah, because the robots still
• 0:14
  don't know who they got money from or who they sent money to
• 0:17
  either, right? Like this is the, you know, this
• 0:20
  is what? So a simple example is address
• 0:22
  poisoning, right? Like if you, if you, you know,
• 0:25
  if you copy a look alike address from your explorer and you know
• 0:28
  from this exploits this, this thing that you really don't know
• 0:32
  who the money came from, right? So, so there's no way to really
• 0:36
  determine. I just have to look at like that
• 0:38
  exact address and like make sure that it's the same address
• 0:41
  you're expecting and, and look at the middle of it.
• 0:43
  Because the address poisoners will take, you know, they'll
• 0:45
  mind for an address that's got the same beginning and end as
• 0:48
  the person that actually that, you know, your friend or whoever
• 0:50
  that actually sent you the money legitimately.
• 0:53
  And then they hope that you copied the wrong address, you
• 0:55
  know, from the explorer, you know, So, you know, robots are
• 1:00
  certainly better at that slightly, right, Because the
• 1:03
  robot can, you know, easily compare the addresses.
• 1:05
  So I, I, I don't think like an AI agents going to get address
• 1:08
  poison because that's a uniquely human error.
• 1:12
  But does the robot really know who the counterparty is?
• 1:16
  Still no, because there's no cryptographic proof of who the
• 1:18
  counterparty is, right? Like anywhere on any chain
• 1:20
  really. So it, it, it, you know, and
• 1:24
  then the robots have to stay private as well.
• 1:26
  Like, you know, it's, it's going to suck when you can, you know,
• 1:29
  figure out like, you know, how much money a robot has and then
• 1:32
  you can go exploit it, right? Because there's, you know, many
• 1:36
  known vulnerabilities and AI agents, you know, prompt
• 1:39
  injection attacks, for example, that are unique to robots.
• 1:42
  So yeah, the robots don't have vulnerabilities that humans
• 1:45
  have, but they have other vulnerabilities that humans
• 1:47
  definitely don't have. They're also fairly easy to
• 1:49
  exploit remotely. So, you know, given that, you
• 1:52
  know, for example, one of our partners is 0 Gai, you know,
• 1:55
  very, you know, high volume, high throughput, you know, very
• 1:59
  cool project that's doing a lot of AI infra, right.
• 2:02
  And so we've actually, they're one of our big strategic
• 2:05
  investors in American Fortress and we're, we're going to be
• 2:07
  deploying our, our privacy tech to their chain first.
• 2:12
  Actually the, our full stack, they're getting it first.
• 2:16
  And it's exactly because of this sort of question.
• 2:18
  We're like, OK, well, if it's a problem with just people doing
• 2:22
  millions of transactions, what happens when you got agents
• 2:24
  doing billions of transactions, right?
• 2:27
  Because, because then a mistake can, you know, really get
• 2:30
  amplified very quickly. And it's not, you know, given
• 2:33
  that the robot's going to catch it or the robot's going to be
• 2:36
  like, you know, fully understanding it, the security
• 2:38
  implications of what it's doing. Because ultimately it's just
• 2:41
  another programme. Sure, a, you know, clever
• 2:44
  programme, or more clever than we're used to in the past.
• 2:46
  But yeah, for sure the AI scaling issue will create more
• 2:52
  problems in the space. Absolutely.
• 2:55
  With privacy, yeah. And the robots need privacy for
• 2:58
  sure, just as much as people. Yeah.
• 3:00
  I mean what happened is few years ago it was with PayPal and
• 3:04
  Apple offering you anonymous linkages or anonymous emails
• 3:08
  with which you can still make a payment and still protect your
• 3:11
  identity. But then with AI, you can do all
• 3:14
  that in, you would say, in a huge scale or a very different
• 3:18
  scale. And then there's no solution.
• 3:20
  And then we are talking about hiding your actual name,
• 3:25
  protecting your identity. And then sometimes we hear this
• 3:28
  concept called the stealth address.
• 3:31
  I don't know if you would like to talk about it.
• 3:34
  Yeah, so in our world so. So here's what a stealth address
• 3:38
  is. So first of all, a stealth
• 3:39
  address can exist on any chain. It's just a mod, a slight
• 3:43
  modification to how addresses are computed, but doesn't break
• 3:47
  the chain. So you know, we've had the
• 3:49
  notion of stealth addresses for Bitcoin since bit 47, which is
• 3:52
  written by Justice Ranvir. He's actually our chief
• 3:54
  architect all the way back in 2015, right.
• 3:58
  So this was pay codes for Bitcoin.
• 3:59
  So the way that worked is a user would generate A payco which was
• 4:04
  a hardened X pub of their private.
• 4:06
  So you know, typically when you start a seed.
• 4:08
  So just for you that don't know, you have a 12 word seed, 15
• 4:11
  words, eighteen word seed, right?
• 4:12
  So that's essentially a, a backup of your private key,
• 4:15
  right? And the private key is a number
• 4:18
  from the private key, you derive A corresponding public key.
• 4:20
  And then you can now derive child public keys and child
• 4:24
  private keys. And this is done very commonly
• 4:27
  in crypto. It's called HD derivation.
• 4:29
  It's in bit 44, BIP 39, etcetera, BIP 32.
• 4:33
  So these are all the sort of original standards and these
• 4:36
  things generally work the same on all chains.
• 4:39
  So like actually, you know, the Etherium address computation
• 4:41
  standard is identical really to Bitcoin other than it's an
• 4:44
  account model versus the ETXO model.
• 4:46
  So the spends work differently, but the address computation is
• 4:50
  actually identical. The Ethereum doesn't swore multi
• 4:54
  set, but I digress. And in any case, this idea of
• 4:58
  stealth address is just been around for a for a while.
• 5:02
  What it does is it lets two parties compute a unique address
• 5:06
  between them, right? So like if I'm sending money to
• 5:09
  you and I'm using stealth address tech, you're going to
• 5:12
  give me my pay code, which is a derivation of your public key.
• 5:15
  And then based on that, I'm going to do some calculations
• 5:17
  and I'm going to find an address on your curve that I can compute
• 5:21
  that you can spend, but only me and you know about that address.
• 5:25
  So yeah. And then when I send money to
• 5:28
  that address, yes, that transactions on chain, but no
• 5:33
  one knows that it's me and you, right?
• 5:35
  So, so, so, So what this does is it means that, you know, every
• 5:40
  time we transact A UTXO, we're actually using a new address.
• 5:44
  We can also perpetually use the first stealth address and still
• 5:48
  meet unique between me and you. And there's no way to reverse
• 5:50
  compute that, right? So there's no way to like
• 5:53
  reverse compute the address like which hardened ex pubs which pay
• 5:58
  codes were used in the transaction.
• 6:01
  So that creates really good privacy already a good privacy
• 6:05
  letter on all chains and tokens. To catch the full discussion,
• 6:08
  head to YouTube, Spotify or Apple Podcasts and search fit
• 6:12
  for privacy. Links are in the description.
• 6:15
  Thanks for listening.