0:00 This is an introduction to KERI, the Key Event Receipt Infrastructure.
0:08 My name is Samuel Smith and I am the inventor of KERI. The reason I invented
0:13 KERI is because the internet is broken and I wanted to solve it. And so the problem
0:19 that KERI solves is called the universal secure attribution problem. Okay, what do
0:24 we mean by universal secure attribution problem? Well, universal means everywhere.
0:30 We want it to be solved everywhere. And secure means do it in a way that can't be
0:39 hacked or forged or impersonated or attacked in the normal way that we get
0:47 attacked on the internet. Attribution means assigning a communication to its
0:53 source. We want to attribute where something came from to its source. We know
0:58 who's sending it to us. And so the problem is, how do we do that everywhere on the
1:04 internet in a secure way so that we know who's sending us information? So that
1:09 doesn't mean we want to enable anybody to track everybody's conversations. We want
1:16 to do it in a privacy protecting way. So that poses additional challenges. But
1:23 first we have to get the security down. Many people approach this problem and they
1:28 sacrifice security. And if you sacrifice the security, then you don't have privacy
1:33 either. So, what we want to do is solve all of these problems at the same time in
1:37 a very integrated, elegant, simple way. Let's talk about what it means. Secure
1:45 attribution is another way of saying we want to establish the authorship of data,
1:51 any and all data sent over the internet, any and all documents, things like
1:55 credentials, entitlements, authorizations. We want to be able to say that we can
2:02 securely attribute who's doing what in a privacy-preserving way. And I keep saying
2:08 privacy because I don't want to give the false impression that we're trying to
2:12 build this universal tracking system. The hard problem is to be able to do this all
2:18 in a way that keeps confidentiality and privacy along with secure attribution.
2:26 That's always been the hard problem. So to restate, the problem is we want to have
2:32 verifiable secure attribution of any communications to its source. Now I added
2:38 the word verifiable. What does verifiable mean? It means that if we're receiving
2:42 information, we need some way to verify where it came from. When I say verify, it
2:48 means in a cryptographic sense. It means doing it without having to trust any third
2:54 party. I don't want to have to trust somebody else to verify what somebody that
3:00 I want to communicate with in a private confidential way says. I need to be able
3:05 to do it by myself. If I have to trust someone, then the security of whoever I'm
3:11 trusting now becomes something I have to worry about. And that is a problem.
3:17 Another way of looking at secure attribution is to say that we want to have
3:23 provenance of any data by anyone, to anyone, from anyone. And so a way we can
3:32 describe that is as authentic data. And then we want authentic data provenance.
3:37 Authentic has two meanings. It means like true to yourself, like you're being
3:42 authentic. That's not the meaning of authentic we're using here. We're meaning
3:46 authentic as you can attribute it to its source. Who painted that painting is an
3:52 authentic painting, right? As opposed to when something's not authentic. It means
3:57 it was forged or somebody was impersonated. Somebody sends you
4:03 information and they say it came from Joe when it really came from Sue. And so the
4:10 whole problem is back to the same thing. We need to securely attribute the source
4:16 of information. So what do I mean by provenance? We often hear the term
4:21 provenance when we talk about like a painting, like fine art, right? Was it
4:26 painted by Rembrandt? And how do you know it was painted by Rembrandt? Well, one way
4:31 that you would provenance it is that you would establish a chain of custody of that
4:38 painting from Rembrandt to whoever Rembrandt sold or gave the painting to
4:43 whoever sold or gave the painting and so on. And if you can follow that chain of
4:49 custody to the present owner or holder of that painting, then you can say, yeah, it
4:54 came from Rembrandt. So if we can solve data provenance, then we can solve many of
5:02 the security problems on the internet because most of them come from the fact
5:08 that we can't solve data provenance. We don't have universal secure attribution.
5:14 And because we don't have universal secure attribution, people can forge things.
5:19 People can impersonate things. Well, when we send data over the internet, we have a
5:24 similar problem. Somebody sends a message. It gets sent to some server in the cloud
5:30 who then sends it to another server, who routes it to another server, who routes it
5:34 to another server, and then it ends up in your inbox. And you go, did it come from
5:40 where it said it came from? Anybody that was in the path could have intercepted it,
5:46 spoofed it, done something to it, and then it shows up in your box and it says, oh
5:50 yeah, this came from Sue, but it really came from Bob. So provenance is about
5:56 establishing secure attribution even though the data traverses multiple hosts
6:04 or servers on the internet. It didn't come directly to us. It wasn't given to us in
6:08 person. It came over the internet and the internet is public. The internet is public
6:13 infrastructure. It's public, which means that lots of people have the ability to
6:18 see what you're sending or to change what you're sending or to mess with it. And
6:22 many of the attacks on the internet come from the fact that it's public
6:25 infrastructure and therefore publicly accessible for attackers to attack. So we
6:31 want to solve this secure attribution problem. Let's dig a little deeper. On the
6:35 internet, we're just sending information. Data is information. It's just bits. Now
6:41 we talked about chain of custody. When we normally think of chain of custody, we
6:45 think of a physical item. So I'll call it those its. They're atoms. So if I actually
6:50 have a painting, there's only one of them. And it takes a lot of work to copy that
6:54 painting. And if I have two, then I have a problem because somebody made a copy. How
7:00 do you tell the original from the copy? Well, if the copy is done really, really
7:05 well, just looking at the two, you can't tell the original from the copy. That's
7:09 why you have to have a chain of custody. You say the original was in possession of
7:14 somebody and they gave it to somebody else and they gave it to somebody else. And the
7:20 copy was never in their possession. possession. So you know it's a copy
7:24 because it wasn't part of the chain of custody. That works for atoms. It doesn't
7:29 work for information bits. So we have a harder problem to provenance data
7:34 information. So we have to use cryptography to do that. One of the first
7:38 things we do with cryptography is that we give an identifier to the source. The
7:45 source now has an identifier. That identifier is an electronic identifier.
7:50 It's more information. It's more bits. To make the ability to security attribute to
7:56 a source, that identifier needs to be unique. And in order to have universal
8:02 secure attribution, that identifier needs to be universally unique. There can only
8:07 be one in the whole universe. The only way that we can guarantee your universally
8:13 unique identifiers is to use cryptography or to use some trusted third party who
8:21 creates them. Well, we don't want to have to rely on a trusted third party, so the
8:27 identifiers must be cryptographic in nature. So, let's suppose that we have a
8:31 system for generating cryptographic identifiers. Now, we must be able to prove
8:38 who controls that identifier, and we have to do it in a verifiable way without
8:42 trusting the third party. If I have a cryptographic identifier and a way to
8:48 verifiably prove controller over that identifier, then I have what is called a
8:54 root of trust. I have the ability now to attribute data to that identifier where
9:01 that identifier is the source of that data. So, I now have constructed the basic
9:06 problem. The way that I can control an identifier or prove control of an
9:13 identifier is to use what we call cryptographic keys. And the specific type
9:17 of key that we use belongs to something called a public key pair or a public
9:24 private key pair where I have a public key and a private key. That is going to be in
9:30 a completely new subject to talk about the basics of public private key pairs. But
9:37 what we've done is set up the core idea of how we get to universally secure
9:41 attribution. And we haven't even really talked about how KERI does it, but I've
9:46 given you an introduction to the hard problem. The hard problem is universal
9:51 secure attribution. Follow along to learn more. Leave any questions in the comments.
9:57 Thank you.
10:27 Thank you.