Hashgraph @ TechCrunch Disrupt – Showcase Stage

September 19, 2017

Watch on YouTube

[Slide: Swirlds Logo]
Andrew Masanto:
Hello Everyone, we are Swirlds.
This is Mance Harmon, the CEO of Swirlds
We have Leemon Baird who isn’t with us, he is the inventor of the technology
He actually holds the fastest ever phd attained at Carnegie Mellon, computer science
And this is jordan fried.
 
[Slide: What is a Hashgraph?]
What swirlds is is the basis of a new technology called the Hashgraph, which we consider to be the future of blockchain technology
What is it? It’s not a blockchain
It’s basically a revolutionary new consensus algorithm which is a complete data alternative to the blockchain
Its mathematically proven to be fast, fair, and what we call asynchronous byzantine fault tolerant
And its implemented in software, it actually works now, it’s not vaporware
 
[Slide: Proof-of-Work Based]
Mance Harmon:
What I am going to do is describe how Hashgraph compares to the rest market, both in terms of performance and in terms of security.
 
So everybody understands what Bitcoin, Blockchain is.
It is proof of work based system.
The architecture is pretty simple.
You have all the users submitting their transactions into the network
These are the miners, these represent the miners.
The miners collect all the transactions, the miners then determine which transaction goes into a block, and if they go into a block, the order within the block.
The miners compete for the opportunity to publish a block to the network, and if they win a hard cryptographic puzzle, they publish their block, the community accepts that block, and puts it on top of the chain.
The problem with this is that it’s extremely slow, the system is by design, slows down the process to give the miners time to come to agreement on which block to put into the network.
For that reason, it can only process single digits of transactions per second.
Also the miner that wins the competition, can actually influence the order of the transactions, and so in some sense, it’s not fair.
The miners determine which transactions go into the network, and if they include that transaction in the block, they determine the order of transactions within the block. So thats a problem.
 
[Slide: Leader-Based]
So to address that problem, the market sort of split into two halves
There is the public side of the network, which are the bitcoin and Ethereum’s of the world
The permissioned or private side of the network uses a different technology altogether
They replaced the inefficiencies of blockchain with old school consensus algorithms like paxos, raft, and pBFT.
The problem here is that there is a leader.
These individual computers represent the members of the network.
The members send all their transactions to the leader
The leader then has the responsibility of putting those transactions in a order, and sending the transactions back to all the nodes – the members of the network.
Well, there is a bottleneck in the leader.
So this can only achieve a 1000 transactions per second, not the hundreds of thousands that are needed.
Secondly, if there is a leader, there is a security vulnerability.
Its trivially easy to execute a DDos attack on the leader.
Every member of the network knows the IP address of the leader.
And so if you have a bank that gets compromised – either through a malicious insider, a disgruntled employee, or a virus, or malware, and the attacker can direct a botnet, then the botnet can attack one computer, the leader, and bring down the entire network.
Nobody talks about at this, but it’s a serious security flaw in the architecture
Well, what will happen is that the leader will time out, and a new leader will be elected.
The problem is that by design, all the members have to know the IP address of the leader, and you can’t detect which member is compromised.
They can do this attack in full anonymity, so you can just follow the leader, and change the target of the botnet attack.
It’s not fair – the leader can influence the order of the transactions.
So you would never build a distributed stock market, for example, on this type of technology, because you could bribe the leader to unfair advantage
Or they could just drop the transaction entirely.
 
[Slide: Economy-Based]
On the public side, again, to address the problems and inefficiencies of blockchain, the market has sort of moved in the direction of creating simple models of economies
Simple policies that incentivize the participants to try and do – to make money.
The assumption is that all the participants will act in their own best interest
If I bet a large sum of money on which block is going to be added to the chain, well then I am incenventize to not do something to lose my money.
The problem is that viruses don’t care about your money.
There is no way to actually write a formal proof that a single attacker can’t do extremely bad things to these economy based systems.
For example, if I were to ask you, can you prove to me, mathematically, that the stock market will never crash again?
Well you would say of course not. But the market is full of people acting in their own self interest. Its a chaotic system.
And there is no way to mathematically analyze it and make those types of proofs.
But, if I could tell you, if you follow certain protocol then I can guarantee that no single actor can bring down the market, would that be of value?
Yes.
Those are the types of formal proofs that we want when building these types of systems that are going to process trillions of currencies, right?
 
[Slide: Voting-Based]
Voting based algorithms for existed for 30 years
They have fantastic security properties.
All the security properties that I just described?
They exist – in voting based algorithms. Pure voting based algorithms.
Not the hybrids we sometimes see in today’s market.
The problem with these voting based algorithms is that they require extreme amounts of bandwidth
They are totally inefficient.
Fantastic theoretical properties, they exist in the literature, but there’s no practical implementation of these because of the bandwidth requirements.
If you could use one of these, there is no leader, there is no opportunity to manipulate the order of transactions.
you can’t Ddos the system, but they are ineffective because of the bandwidth requirements
 
So the question is, how do you use a voting based algorithm, without having to cast the votes?
 
[Slide: Hashgraph with Virtual Voting]
That’s what Hashgraph does.
Hashgraph makes it possible to have all the fantastic security properties of voting based algorithms, without ever casting a single vote over the network.
And it works pretty simply.
Basically, if we all have a copy of a database, all the transactions that get gossiped have to go to everybody, so that is the minimum bandwidth, baseline requirement.
We can add a tiny amount of information on top of that baseline.
Basically, when I talk to you, in addition to passing you the transaction, im going to pass you a little bit of information that describes or says who the last person I’ve talk to.
And you are going to do the same.
So every time you talk to a member of the network, you pass the transactions, and you say who it was you spoke to last
Using that information, we can build a picture of the gossip.
Basically the graph, the Hashgraph, of the gossip about gossip.
Who has talked to whom, and when
Using that, as an input to one of these pure voting based algorithms, we can calculate internally, what each member would vote, if they were to have cast their vote over the network.
In other words, we do virtual voting.
We have gossip about gossip that builds up the Hashgraph, and we use the Hashgraph with these pure voting based algorithms, to have virtual voting.
Which means, we can process hundreds of thousands of transactions per second, and we have phenomenal security properties.
We actually solve the problem of making asynchronous byzantine fault tolerance practical.
Which means, all of those problems that you heard yesterday in the general sessions, in how to scale and achieve the solution the solutions we want to achieve.
We have it in Hashgraph.
We have the fundamental building block that enables everything the market is looking for.
 
[Slide: Security]
So in terms of security, we are unique in the market.
Nothing else achieves the level of security that we’ve achieved.
 
[Slide: New Properties – New Use Cases]
In terms of use cases, we do everything blockchain does, and then a whole lot more.
Because of the fairness properties we have, you could build a distributed stock market on this.
Fair distributed stock market.
You could build world of warcraft, a distributed world of warcraft, on top of our technology, that actually works – no servers.
You could build an ebay, a distributed eBay
 
Andrew Masanto:
We actually had a hackathon yesterday, and someone built a distributed ebay.
I don’t know if you guys have seen the ICOs that have happened recently, but someone built a distributed ebay on Hashgraph in 24 hours
A distributed game, all sorts of crazy, crazy stuff.
And that’s just the start of the use cases you get when you have properties like Hashgraph has.
 
[Slide: Additional Info]
Mance Harmon:
Alright, that it
This is where you can get more information.
You can go to Hashgraph.com you can go to swirlds.com we believe that in the future, blockchain will be called Hashgraph.
 
Jordan Fried:
And we also have a table over here, with all the sponsors on the left hand side, come talk to us.
We are happy to go into detail about what we are doing and how you can deploy in your business.