Cryptography Reference
In-Depth Information
on 64 sufficiently big films — one film for each bit position — we can create
all possible ciphertexts from a known plaintext in many — but a finite number
of — steps, again distributed across 64 films, i.e., one film for each bit position.
We link these bit levels photographically such that a 0-bit can be only where
the eavesdropped ciphertext was created. The position of this (theoretically)
single 'light spot' reveals the DES key.
What do we need the spray can for? Well, we can use it to create random
bits: we simply spray color on a film substrate such that about 50 % of the
dots appear black. We can use the result to generate the 64 bit levels of all
possible keys.
This is just a very simplified description of the method. Interested readers will
find the details on our Web site or in the Proceedings of EUROCRYPT '98.
However, knowledgeable people will already have identified several technical
problems:
•
Shamir assumes a technically feasible storage density of about one billion
bits per square inch. This means that one bit would have an expansion
of 1
µ
m, i.e., we'd need clean-room conditions for film processing.
•
People knowledgeable about chip production processes know of the prob-
lems involved in trying to exactly adjust the films, mainly because of their
deformation (due to humidity, temperature, and inhomogeneity). This
means that we'd have to work with much smaller sections and many
adjustment marks and many single steps per exposure.
•
Superimposing the films so they can act as exposure masks isn't as easy
as it may sound: the light source has to supply parallel beams, and the
light diffusion has to remain very small. Interference effects must be
expected.
•
Photographic layers are not homogeneous; instead, they can have 'holes'
(small areas without light-sensitive particles).
•
The spraying process works (if at all) only provided the drops are not
electrically charged; otherwise, there will be a minimum distance between
drops.
It appears more realistic to use small squares with an edge length of 5 to
10
µ
m. Also, it is by no means clear whether or not the financial cost for a
visual cryptanalysis will remain within the acceptable range at a reasonable
