Cryptography Reference
In-Depth Information
following concept would probably have been better: the user sees a 128-bit
key, but the key space comprises only 2
40
elements due to secret dependencies
between the bits. These dependencies would be known only to the manu-
facturer, and surely to the NSA. However, the question is how long such a
dependency could be kept secret. One would have to implement it in a crypto-
logically secure and variable way. This would certainly be a challenging task
for research.
How much would it cost to brute-force break such a key? Suppose a very fast
implementation of the algorithm (e.g., in assembler language) on a Pentium chip
manages the decryption of one million plaintexts in one second. 2
40
keys cor-
respond to about 10
12
possibilities, i.e., the Pentium Pro chip would take about
500 000 seconds on average. That's almost 6 days. Using the time - memory
tradeoff and several computers, the cost can be arbitrarily reduced. In short,
this key space is already much too small for current and privately available
computer technology, let alone special hardware. The only protection could
be the secret bit dependencies within an externally longer key, as mentioned
above.
Netscape didn't use such dependencies, and what's more, in software that was
supposed to handle credit card numbers. However, the attacks that became
known were much less clever than one would expect: hackers stole credit card
numbers directly from Web servers, since they had been lying around there with
almost no protection in place. This shows impressively that good cryptography
is only part of comprehensive security.
Toward the end of the 1990s, even politicians and intelligence agencies appar-
ently realized that secure cryptography 'cannot be avoided'. I assume that the
business world had its share in arriving at this realization. Export regulations
changed so quickly that I couldn't keep up, and I have to refer you to the
Internet to learn the current status. Meanwhile, products using 128-bit keys
may be exported from the USA within corresponding approval procedures. It
is probably more rewarding to turn one's interest away from algorithms and
key lengths and instead to look at backdoors built in application software, or
better yet, to automated hacker techniques.
There has also been much speculation about the
Wassenaar accord
that
includes export regulations for 33 countries, including the EU countries. As
usual, the last meeting of the Wassenaar countries took place at a secret loca-
tion near Vienna, Austria, and we know accordingly little about the agreements
made there. Rumors have it that public-domain software was not concerned.
