Re: One time pads

[Bill Stewart]

At 10:52 PM 10/17/2002 -0700, Morlock Elloi wrote:

 I have a working OTP system on $40 64 Mb USB flash disk on my keychain.

 Cute.  Is it available?

$39 + tax in Fry's.


I don't mean the disk - there are lots of those.
I mean your software.
Also, can your tool use floppies instead of USB keys?
There are problems with KGB-quality attackers recovering overwritten data
which are probably much more serious for disks than flash rom,
but they're nearly universal and good shredders work well on them.


 How do you prevent other applications from reading the file off your
 USB disk, either while your application is using it or some other time?

I don't care. No one knows about it enough to set a trap in a random PC 
(and if
They do we're in deep shit anyway.) This is the reason for not releasing the
(trivial) program. Write your own and let it be your group key ... say, 
40-bits worth ?

USB key disks look like an obvious target for eavesdropping in general.
(They're also the best medium for re-inventing the floppy-disk virus:-)


 Since you say that Used bits are securely deleted,
 does your application distinguish between using the pad to encrypt
 and using the pad to decrypt (which are basically the same thing,
 except for destroying the key bits the second time)?

You destroy bits *every* time. The routine that reads bits overwrites them.
Messages are fixed size, index into OTP file is a part of the message, each
user gets starting offset assigned to avoid synching problems.


You need to use each bit twice - once to encrypt, and once to decrypt.
Destroying them after the first use is a bad idea

Re: One time pads

[Bill Stewart]

At 02:04 PM 10/17/2002 +0200, Eugen Leitl wrote:

It is important to note that currently NMR bases systems only allow for
6 qubits. Only very recently we're getting practical qubits in solid state.
.
Everybody realizes that we're discussing currently completely theoretical
vulnerabilities, right?


Of course.  But without quantum computing, you can do computations on
your basic cheap computers that are secure against crackers for the
expected remaining life of the universe, so your threat models are
much more controllable.  Obviously you still need to worry about tempest,
computer viruses, cameras in the ceiling, and rubber hose cryptanalysis,
but threat models that just involve someone intercepting your message
aren't a problem.

Quantum computing is the one thing that anybody's thought of that
has a mathematically possible chance of breaking that.