6 ms·

whats to stop the node replying with the hash of the file saying "yes the file is here" when in reality no file is there

ZitchDog 12y ago
Haven't read the paper, but could you ask for a hash of a random offset+byte count of the file?
- slipstream- 12y ago
  Exactly what I was thinking.
- wyager 12y ago
  That only works on direct person-to-person systems, not decentralized systems where the entire network has to reach consensus.
_prometheus 12y ago
the proof-of-retrievability. see http://cseweb.ucsd.edu/~hovav/dist/verstore.pdf http://cseweb.ucsd.edu/~hovav/dist/verstore.pdf
- philsnow 12y ago
  I'm at work and don't want to read either of the papers at the moment, but if A asks B to provide proof of having stored some bytes, could B proxy that request to C and replay the reply to A ?
  - justin_ 12y ago
    As long as a node you control has the data and can retrieve the file, I don't think there's a problem. If you could proxy the "proof of retrieval" than you could surely proxy the retrieval itself. Pooling resources shouldn't break anything.
    - sp332 12y ago
      But you could make it look like several computers have copies of the data when there's really only one copy of the data. You would get credit/coins for multiple copies and reduce the redundancy of the files.
      - anonymousDan 12y ago
        I haven't RTFA or any of the papers but couldn't you perhaps add a different nonce to each copy of the file you want stored before encrypting?
      - justin_ 12y ago
        From what I read in the paper, it sounds like you _don't_ get paid for providing the data. You get paid for proving you have it when mining a block. Distributing your storage across multiples nodes would simply make it easier to "mine", but wouldn't get coins passively. That is, if I'm understanding the protocol correctly.
        _prometheus 12y ago
        You're right that nodes get paid for proving. Though you do also get paid for providing pieces in Get transactions. On outsourceability, Filecoin today makes no effort against it. But see Permacoin for a great (and compatible) solution.
  - nullc 12y ago
    FWIW, thats usually called 'delegation' or 'outsourcing'. Apparently it's intentionally not resistant to this (sounds like a pretty bad idea— IMO, but apparently intentional). See the IRC log I linked above.
- padelt 12y ago
  FWIW: For the simpler case of two single entities there is a non-probabilistic (whole data needed for proof) proof-of-retrievability concept: https://www.researchgate.net/publication/4326385_Reliable_Evidence_of_Data_Integrity_from_an_Untrusted_Storage_Service https://www.researchgate.net/publication/4326385_Reliable_Ev... Disclaimer: I am one of the authors.

ZitchDog 12y ago

Haven't read the paper, but could you ask for a hash of a random offset+byte count of the file?

slipstream- 12y ago
Exactly what I was thinking.
wyager 12y ago
That only works on direct person-to-person systems, not decentralized systems where the entire network has to reach consensus.

_prometheus 12y ago

the proof-of-retrievability. see http://cseweb.ucsd.edu/~hovav/dist/verstore.pdf http://cseweb.ucsd.edu/~hovav/dist/verstore.pdf

philsnow 12y ago
I'm at work and don't want to read either of the papers at the moment, but if A asks B to provide proof of having stored some bytes, could B proxy that request to C and replay the reply to A ?
- justin_ 12y ago
  As long as a node you control has the data and can retrieve the file, I don't think there's a problem. If you could proxy the "proof of retrieval" than you could surely proxy the retrieval itself. Pooling resources shouldn't break anything.
  - sp332 12y ago
    But you could make it look like several computers have copies of the data when there's really only one copy of the data. You would get credit/coins for multiple copies and reduce the redundancy of the files.
    - anonymousDan 12y ago
      I haven't RTFA or any of the papers but couldn't you perhaps add a different nonce to each copy of the file you want stored before encrypting?
    - justin_ 12y ago
      From what I read in the paper, it sounds like you _don't_ get paid for providing the data. You get paid for proving you have it when mining a block. Distributing your storage across multiples nodes would simply make it easier to "mine", but wouldn't get coins passively. That is, if I'm understanding the protocol correctly.
      - _prometheus 12y ago
        You're right that nodes get paid for proving. Though you do also get paid for providing pieces in Get transactions. On outsourceability, Filecoin today makes no effort against it. But see Permacoin for a great (and compatible) solution.
- nullc 12y ago
  FWIW, thats usually called 'delegation' or 'outsourcing'. Apparently it's intentionally not resistant to this (sounds like a pretty bad idea— IMO, but apparently intentional). See the IRC log I linked above.
padelt 12y ago
FWIW: For the simpler case of two single entities there is a non-probabilistic (whole data needed for proof) proof-of-retrievability concept: https://www.researchgate.net/publication/4326385_Reliable_Evidence_of_Data_Integrity_from_an_Untrusted_Storage_Service https://www.researchgate.net/publication/4326385_Reliable_Ev... Disclaimer: I am one of the authors.