[9 May 16 @ 6:40 PDT] For those interested in the hash collision attack discussion, it turns out there is a faster way to scan your set to find the collision: you’d keep a sorted list of the hashes for each TX you generate and then use binary search to check that list for a collision for each new TX you randomly generate. Performing these operations can probably be reduced to N lg N complexity, which is doable for N ~2^32. In other words, I now agree that the attack is feasible.
Cheers, Peter hat tip to egs > On May 9, 2016, at 4:37 PM, Peter R via bitcoin-dev > <[email protected]> wrote: > > Greg Maxwell wrote: > >> What are you talking about? You seem profoundly confused here... >> >> I obtain some txouts. I write a transaction spending them in malleable >> form (e.g. sighash single and an op_return output).. then grind the >> extra output to produce different hashes. After doing this 2^32 times >> I am likely to find two which share the same initial 8 bytes of txid. > > [9 May 16 @ 4:30 PDT] > > I’m trying to understand the collision attack that you're explaining to Tom > Zander. > > Mathematica is telling me that if I generated 2^32 random transactions, that > the chances that the initial 64-bits on one of the pairs of transactions is > about 40%. So I am following you up to this point. Indeed, there is a good > chance that a pair of transactions from a set of 2^32 will have a collision > in the first 64 bits. > > But how do you actually find that pair from within your large set? The only > way I can think of is to check if the first 64-bits is equal for every > possible pair until I find it. How many possible pairs are there? > > It is a standard result that there are > > m! / [n! (m-n)!] > > ways of picking n numbers from a set of m numbers, so there are > > (2^32)! / [2! (2^32 - 2)!] ~ 2^63 > > possible pairs in a set of 2^32 transactions. So wouldn’t you have to > perform approximately 2^63 comparisons in order to identify which pair of > transactions are the two that collide? > > Perhaps I made an error or there is a faster way to scan your set to find the > collision. Happy to be corrected… > > Best regards, > Peter > > _______________________________________________ > bitcoin-dev mailing list > [email protected] > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev _______________________________________________ bitcoin-dev mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
