Cryptography Reference
In-Depth Information
blocks
P
1
...P
k
. In the beginning, all of the blocks are initialized to
be empty. Then a random collection of blocks are filled with random
numbers so that there will be random data to be used for the
n −
1
blocks used to store a file.
A keyed random number sequence,
r
1
,r
2
,...
, is still used to select
the parts, but it is modified to take into account the fact that many of
the blocks may hold no data. In this case, the
n −
1
parts used in the
secret sharing algorithm will be the first
n −
1
full
parts identified by
the randomnumber sequence. That is,
P
r
1
and
P
r
2
might be unfilled
yet, but
P
r
4
have random data available either from another
file or from the initial seeding of random information. So we use
those two parts. The final part,
P
r
3
and
F
+
P
r
1
+
P
r
2
+
...
+
P
r
n−1
,isnot
,butatthefirst
empty
part identified by the random
number stream, in this case,
stored at
P
m+1
.
This technique eliminates the need for any table linking
P
r
1
P
m+1
with the name for the file, but it adds other complexity. It is possible
to identify the empty and full blocks when the file is being stored
but this will change as more blocks are added to the mixture. One
solution is to add some additional information to the final part that
includes the name of the file, perhaps encrypted with a key, and
the locations of the
1
part needed to assemble the final part.
The file could then be recovered by computing the random number
stream,
n −
, and testing each part identified by the stream until
the final part is found. The name attached to the final part will make
it possible to identify it and the locations of the
r
1
,r
2
,...
1
parts stored with
that name will make it possible to recover the entire file.
An approach like this is still vulnerable to some simple denial-
of-service attacks like storing a large number of garbage files. This
will make it much more likely that deleting one file will not damage
anything of value.
n−
10.6 Publius
Another popular tool for storing documents securely is the
Publius
system designed by Marc Waldman, Aviel D. Rubin and Lorrie Faith
Cranor.
The systemuses secret sharing algorithms to split up a document
between
pieces are sufficient to re-
cover it. The tool uses a protocol for recovering the document that
rests on top of HTTP making it possible for it to work fairly innocu-
ously with standard webservers. The solution is notable because it
offers the publisher and only the publisher an opportunity to delete
and update the documents after they're distributed.
n
different servers so that any
m
