Moore vs. Crypto

In "Encryption for a Small Planet" (March), Thom Stark neglects to mention the effect of Moore's law on data security. My corollary to Moore's law is this: To maintain the same level of data security, you must increase your key length by 1 bit every year. By Moore's law, processor performance doubles every 18 months; however, there are also advances in algorithms and mathematics, as well as an increasing number of computers available to a would-be cryptanalyst. Since a doubling of computer power is equivalent to shortening the key by 1 bit, a conservative estimate is that the key length should increase by 1 bit every year.

Stark's theoretical estimates for the maximum times needed to crack keys of different lengths are correct in today's terms, but incorrect in absolute terms. In his example, RC-4 with a 56-bit key takes 2691.49 years to crack. A 56-bit key in 16 years is equivalent to a 40-bit key today. Wait 16 years, and then start your cryptanalysis: The 56 bit-key would take 16 years and 15 days to crack. If the U.S. government insists on maintaining its curre nt stance on the export of encryption technology, then it should at least frame future legislation to allow cryptography vendors to increase key lengths by 1 bit every year, with a review of the baseline, say, every 10 years.

Martin Budden
Richmond, Surrey, U.K.
martinjb@cix.compulink.co.uk

There's nothing absolute about Moore's law; it's an observation of a trend, and the trend will slam into the realities of minimum scale and maximum waste heat transfer before too many more microprocessor generations elapse. Your argument about advances in algorithms is a better one; there might be indefinite room for improvement there.

The ground on which the U.S. government maintains its limits on exportable enc ryption strength amounts to "we have to be able to decipher the bad guys' traffic." If you buy that, then there's no reason to ever increase the key-length limit. Of course, strong encryption is available from non-U.S. vendors, and further, if criminals and terrorists don't respect laws against crime and terror, why would anyone expect them to respect laws against exporting strong encryption?--Thom Stark