4 ms·

He/they may be known to some. https://medium.com/cryptomuse/how-the-nsa-caught-satoshi-nakamoto-868affcef595 https://medium.com/cryptomuse/how-the-nsa-caught-sa

He/they may be known to some. https://medium.com/cryptomuse/how-the-nsa-caught-satoshi-nakamoto-868affcef595 https://medium.com/cryptomuse/how-the-nsa-caught-satoshi-nak...

wyxuan 7y ago
Stylometry only works too an extent with helping to identify what's what. I could just stylometry to show that this document was written by a professional screenwriter, but it writing style isn't unique enough to identify one person out of billions. Gwern has an excellent post about this.
- s_y_n_t_a_x 7y ago
  It depends on how much data you have. When you have everyone's emails you can rank everyone by closest match and narrowing parameters.
- varenc 7y ago
  Agree 100%. I'd believe that the DHS has identified Satoshi, but very very skeptical that they could use mass email collection to do it. Stylometry could work though if you limited your search to the set of people active on niche pre-Bitcoin lists/forums where Satoshi was likely to be active under a different name.
- zawerf 7y ago
  I remember a really impressive demo where someone found all the alt accounts of a HN user using stylometry: https://news.ycombinator.com/item?id=17944484 https://news.ycombinator.com/item?id=17944484 The account didn't have that many comments on it (maybe ~1000 karma) He also claims to have used the same algorithm for identifying company insiders for trades (see parent comment in that thread for context). I don't think he was bullshitting even though his tool isn't public because he has a different NLP project that works reasonably well (https://hnprofile.com/ https://hnprofile.com/ which can find users based on their interests).
donaldknuth123 7y ago
The likelihood that they were able to use Stylometry to identify him is near Nil in my opinion. State of the art language models out of Google, Stanford, MIT etc. are barely able to accurately identify authors based on style. I doubt very much that even with "trillions of emails" the NSA was able to do this successfully.

He/they may be known to some. https://medium.com/cryptomuse/how-the-nsa-caught-satoshi-nakamoto-868affcef595 https://medium.com/cryptomuse/how-the-nsa-caught-satoshi-nak...

wyxuan 7y ago

Stylometry only works too an extent with helping to identify what's what. I could just stylometry to show that this document was written by a professional screenwriter, but it writing style isn't unique enough to identify one person out of billions. Gwern has an excellent post about this.

s_y_n_t_a_x 7y ago
It depends on how much data you have. When you have everyone's emails you can rank everyone by closest match and narrowing parameters.
varenc 7y ago
Agree 100%. I'd believe that the DHS has identified Satoshi, but very very skeptical that they could use mass email collection to do it. Stylometry could work though if you limited your search to the set of people active on niche pre-Bitcoin lists/forums where Satoshi was likely to be active under a different name.
zawerf 7y ago
I remember a really impressive demo where someone found all the alt accounts of a HN user using stylometry: https://news.ycombinator.com/item?id=17944484 https://news.ycombinator.com/item?id=17944484 The account didn't have that many comments on it (maybe ~1000 karma) He also claims to have used the same algorithm for identifying company insiders for trades (see parent comment in that thread for context). I don't think he was bullshitting even though his tool isn't public because he has a different NLP project that works reasonably well (https://hnprofile.com/ https://hnprofile.com/ which can find users based on their interests).

donaldknuth123 7y ago

The likelihood that they were able to use Stylometry to identify him is near Nil in my opinion. State of the art language models out of Google, Stanford, MIT etc. are barely able to accurately identify authors based on style. I doubt very much that even with "trillions of emails" the NSA was able to do this successfully.