So, I decided to return to my starting point for all this and look at what happened to MtGox's bitcoin after they "demonstrated control of a vast (well, at least pretty impressive) reserve of coins in 2011 in order to counter suggestions they might not have enough reserves of "real" coin to cover people's trading accounts. It was a bit of a dog and pony show. The new traces show up some very interesting features, including one or two I still find pretty baffling.
So, here is the starting point for the trace, visualised using Gephi. Remember, each sphere is a bitcoin address, and in this case each little arrow is a transfer route that was exercised in the first 5000 transactions after the "show". The MtGox wallet shows up red, a little below centre, with numerous arrows emanating from it. The redder the node, the better it is connected to MtGox. There is quite a lot of "fan-out" even in this image, with bitcoin heading off in lots of directions.
If we pull back a bit we can see this is just part of a wider ecosystem:
Around 5000 addresses - with some noticeable "clumps" of organisation (although nothing on the scale of Satoshi Dice (discussed in a previous post). The group shown below is the biggest.
This one is Instawallet - a "we host your wallet" site that folded stating hacking and loss of coins as the reason. Another of the notable groups is shown below:
The prominent "hub" here is Coinb - another wallet service who also appear to have folded. The presence of these two (defunct) wallet sites does prompt the question as to whether MtGox could have borrowed the bitcoin they used for the "show"? That has to remain a subject of conjecture - at least for now. Other explanations could be that Gox users withdrew funds and stored them in online wallets - or that Gox stored some of their funds there.
Now here is the strangest feature of all. You have to pull right back to see it.
See the big loop? That appears to be a string of hunderds of addresses, with bitcoin be transacted on and on, along the chain. I have no idea what the idea of this is... Is it an obfuscation method? If so, it isn't a great one as it stands out like a beacon of bizarreness. We can zoom in:
Here is the start of the loop. As you can see, the loop has two strands of addresses, which transact along the chain and also direct a little bitcoin in to the "centre" addresses. The second strand appears to be second pass around the loop - so its really a big spiral. Could be if I run more transactions through the trace we'd get more turns. If anyone can explain what's going on, I'd really like to hear.
I'm off for another round of development of the python scripts - I have a couple more additions in the pipeline that will give more detail.






Thank you David for the interesting post!
ReplyDeleteI'm working on blockchain data for my master's thesis and planning to analyze Mt Gox incident, how did you get all the addresses that belong to Mt Gox? Did you get them from blockchain.info?
I already posted a comment but couldn't see it here :(
Hi Israa. Glad you found this post interesting. If you want more details of my analysis, much of it is in some of my earlier posts. The post "Tracing stolen bitcoins wherever they go" has a link to a post I found by another investigator looking at the Gox incident. That gave me addresses and a trail to follow. My post "Unlocking the Blockchain" gives a ref to John Ratcliffe's blog. He wrote the C++ code I customised to make my C++ blockchain trace app. I also use some python code to turn the raw data into files I can feed to Gephi. Generally I use Blockchain info to do follow-up checks on addresses that show up as interesting on my Gephi traces. Message me back if you want more info - I'm happy to give you more details etc. via email. Happy to help. Great topic for further research and so much more to be done. I wish I had more time to dig deeper and try out some more ideas.
DeleteHello David, thanks for the reply and useful info.
DeleteI also customized John Ratcliffe's code to load all the blockchain data into postgres (together with blockchain postgres dump link: http://dumps.webbtc.com/bitcoin/) My plan to use R to analyze Mt Gox coins..
I would be very grateful if you provide me with more details, my email address: ialqassem@masdar.ac.ae
Thanks a lot!
This comment has been removed by the author.
ReplyDeleteHi Israa - thanks for the webbtc link. That's new to me and looks very useful. Currently downloading but will take all night as the files are huge. My modified version of John Ratcliffe's code stops it needing to cache every transaction in memory (which didn't work on my machine) - instead it just follows an expanding trail out from a specified start address.
ReplyDeleteI'll send you a trace from the MtFox incident and my python script that converts this to gexf format for import into Gephi. Good luck with the research!
Thank you so much David!! So nice of you :))
DeleteIf you want to speed up the installation use Internet Download Manager..
And I'll send you the instructions on how to setup postgres and the blockchain database.