‘‘SAFE is a network made up of the spare computing resources of its users. It can be thought of as a crowd-sourced Internet. When you contribute to the network, you allow it to store encrypted pieces (not complete files) of other people’s files onto your computer. As you turn your computer off all network data is removed from your machine and will start to fill up again when you turn it back on.’’
So forgive my ignorance but if a machines hard drive fails and the ‘SAFE’ information it was storing is lost before it can be removed…what happens in that scenario …?.
Thank you for your quick reply …how many times is the same chunk of information duplicated on separate machines…?..I’m just trying to ascertain just how secure specific pieces of information actually are.
Thank you digipl…for your concise answer…so from that I’m assuming the security of the whole SAFE network depends on enough computers using it to provide enough copies to spread out the risk.
A few components at play here, Security or consensus is a group of nodes close to something who all agree on an answer to a particular question (i.e. does a safecoin live here? who owns it etc.). Then security of individual data items, again group secured Structured Data is signed by owners and replicated across a group. Immutable data, is replicated across 3 groups who all hold 2 live copies at any time.
Security of data at rest is self_encryption (see code and papers) and in transit via several methods depending on where
XSalsa20 stream cipher with Poly1305 MAC authentication (symmetric)
X25519 Keys XSalsa20 stream cipher (encrypt) with Poly1305 MAC authentication (asymmetric), we also use anonymous version of this as well (unauthenticated)
I think you are thinking of the security of losing 1 chunk being the number of these groups and to a great extent it is. The 3 groups holding the chunk (ignoring local and cache copies) will be distributed evenly across the address range (and geographically). The holders of the data are directly connected to their closest group. On connection loss a copy is made immediately by the group. So should be quite quickly backed up. The chance of all 6 copies going off line at a time very close together is chosen to be infeasible (like a normal kad network with group size of 8, which states 8 nodes going off in a refresh period (1 hour) is infeasible).