If You’re Smart Leave the room right now “Bitcoin turns nerd forums into libertarian forums” This is true Bitcoin is a particularly effective DoS against security professionals Why?
Security Inversion Normal Code Looks like it might be OK up front Scratch the surface, it’s actually really bad BitCoin Looks really bad up front Scratch the surface, it’s actually surprisingly good We aren’t used to systems with these characteristics This code has the mark of having been audited by People Like Us And quants
The basic summary BitCoin is absolutely not anonymous BitCoin clearly does not scale In the long term It does work for now though This isn’t 0day stuff, this is basically declared almost entirely up front
What Is BitCoin A really strange use of cryptography “Strange” is not a sufficient, interesting, or even vaguely competent way to mark a system as insecure It’s a decent way to say “this is not the normal way things are put together” Two systems mated together A peer to peer network that does a best case effort to synchronize data (loose “transactions” and solved “blocks”) across as many nodes as possible A Chinese Lottery that canonicalizes subsets of synchronized data, using the difficulty of finding partial hash collisions
The Basic Idea (In A Nutshell) 1) I’m hearing about all these transactions going on – Alice is paying Bob, Bob is paying Charlie, etc 2) I hash all the transactions I’ve heard about, with some random information, and the hash of the last time someone did that, until there’s a partial collision First n bits equals 0 N is automatically determined based on how hard it has to be for one block to be found about every 10 minutes This is a block 3) I send everyone my “block” – transactions plus hash of previous block plus random data. This gives me 50 bitcoins (for now). 4) I can now “sign over” those bitcoins, from my private key, to other people’s (or my) public key. 5) Repeat until there’s lots of people with lots of BitCoins Possibly purchased instead of “mined”
Interesting Traits The basic concept is actually relatively solid Assuming partial collisions are predictably hard to find Assuming ECDSA works Basic Idea 1: Money can’t be created from nothing – hashing is needed Basic Idea 2: Transactions can’t be blocked or reversed by a central entity – “is none” It makes security engineers talk like monetary scientists That’s sort of OK, economists pretend to do that too… Seriously, that’s silly– lets just talk tech, OK?
Epic Scalability Quote 1(https://en.bitcoin.it/wiki/Scalability) “The core BitCoin network can scale to very high transaction rates assuming a distributed version of the node software is built. This would not be very complicated.” Because there’s nothing easier to do, than make a system distributed This is totally not one of the Hard Problems Of Computer Science By “Distributed” they mean “Centralized” WhyBitCoin is uniquely hard to audit It claims the advantages of its present architecture, and its future architecture, while rebutting the disadvantages of one with the advantages of the other Instead of saying, “We don’t do that”, they say “Something else could do that”
Scalability Costs: Network Bandwidth “Let's assume an average rate of 2000tps, so just VISA…. Shifting 60 gigabytes of data in, say, 60 seconds means an average rate of 1 gigabyte per second, or 8 gigabits per second.” :O
Up and Down Going up “Let's take 4,000 tps as starting goal. Obviously if we want BitCoin to scale to all economic transactions worldwide, including cash, it'd be a lot higher than that, perhaps more in the region of a few hundred thousand transactions/sec.” And the need to be able to withstand DoS attacks (which VISA does not have to deal with) implies we would want to scale far beyond the standard peak rates. TB/sec Going down Even at 1/100th of VISA, that’s still 10MB/sec
Are There Future Optimizations? “Because nodes are very likely to have already seen a transaction when it was first broadcast, this means the size of a block to download would be trivial (80 bytes + 32 bytes per transaction). If a node didn't see a transaction broadcast, it can ask the connected node to provide it.” Potential 50% savings! Could go from 1GB to 500MB/sec
What About Storage? In order to validate a transaction, you need all blocks up to the present one Joining BitCoin today == downloading 200+MB history all the way to the start of time That only increases “ A 3 terabyte hard disk costs less than $200 today and will be cheaper still in future, so you'd need one such disk for every 21 days of operation (at 1gb per block).” So you get to participate directly in BitCoin, at the low low cost of $200 a month Assuming zero costs of running a storage array
CPU? ”A network node capable of keeping up with VISA would need roughly 50 cores + whatever is used for mining (done by separate machines/GPUs).” In the long run, that’s what it takes to participate (assuming no DoS, which would take 5000 cores) (You actually need to validate all historical transactions too)
OK, so you end up with supernodes and normal nodes What are the characteristics of supernodes? They’re banks “Welcome to the new boss, who looks suspiciously like the old boss” I’m not saying banks are bad or anything The “peer to peer” model of BitCoin eventually goes away; as soon as the thing gets big, the entire thing switches to a banking model
Reality of Banking As the network gets bigger, fewer and fewer nodes can be banks Only so many parties can exchange a gigabyte a second. The 50% threshold is inevitable BitCoin banks still can’t gin up money BitCoin banks can’t forcibly take money Unless they hold the private keys for the user, which they might BitCoin banks can refuse to accept blocks with “undesirable” transactions Don’t need 50% -- just need enough to inconvenience 50% to accept your opinion Can block undesirable transactions Can recompute blocks w/o certain transactions (reversal) This offers a host of ugly semantics
Already Suffering This BitCoin’s security model is base on the idea that nobody can control more than 50% of the network Exact PetaFlop count unclear, but >40 and <200 Weird metric, given that crypto uses integer operations when FLOPS are floating point Several times more than largest supercomputer Pools are breaking this #1 pool has 41% #2 pool has 30% “Security through ostracism” to Pitchfork Security DDoS against #1 pool
Bad Choice Of Hash Standard Existing model can be accelerated massively with GPUs Just 2x SHA-256 Could have been bcrypt or the like, in which performance does not scale with pure processing speed Basically adds memory and serialization dependencies Wasn’t implemented, so now we have shortages of GPUs…
What About Anonymity? The full worldwide transaction history is stored and shared, forever and ever Everyone has names like: 1MQbbWUi2scKdZ4KtMMSUSvVmxi6XtEeaC How do you know who you’re paying? You don’t Everyone is encouraged to make up new names for every transaction Actually how you can tell why someone is paying you Out of band, you tell someone “to pay me, pay this address” When that address is paid, you can dereference to your own private transaction Do lots of random names equal anonymity?
Names Are Linkable (see blockexplorer.com) All FROM sources are effectively the same person (or linked IDs) Almost all TO destinations are payee and payor
Reality of Anonymity As BitCoin “fights fragmentation”, it merges identities As it merges identities, it…well, merges identities There are other models of using BitCoin in which money goes in, stays, and then presumably goes back out Again, it’s amazing how much this looks like a bank. Not saying banks are bad, just don’t tell me BitCoin doesn’t morph into the banking system
So, with this all being said BitCoin is working, today That counts for a lot It will not work this way forever It will not have today’s security properties forever If you define the loss of today’s properties a serious loss of value, then there are Ponzi-ish characteristics in plain view I’m not going to make that claim, however
Conclusion This was just a quick summary BitCoin is actually well designed, if you accept that anonymity and scaling forces the entire present model to be shifted into something that effectively looks like banking I’ll talk about more another time