back to article 'Sandwich attack' busts new cellphone crypto

A new encryption scheme for protecting 3G phone networks hasn't even gone into commercial use and already cryptographers have cracked it - at least theoretically. In a paper published Tuesday, the cryptographers showed that the Kasumi cipher, which is also referred to as A5/3, can be broken using what's known as a related-key …


This topic is closed for new posts.
  1. Anonymous Coward

    I don't think much of their sandwich recipe!

    >"two thick slices at the top and bottom and a thin slice in the middle."

    I *knew* Dagwood Bumstead, and you, sir, are no Dagwood Bumstead!

  2. Tom Chiverton 1

    Ya see

    Ya see, this is why you must not do security in secret. The clever people don't get to have a go at it till it is too late for you.

  3. AlistairJ

    k'sue you

    Solution is to replace keys regularly, like in any well-managed crypto system.

  4. John Smith 19 Gold badge

    Adi Shamir

    Is he also the guy who heads the company that supply the crypto for Sky Digital?

    Angling for a contract for the replacement?

    But note 2^38 plaintexts? That sounds like *quite* a lot.

  5. Pitpat

    2^32 time

    Errrm. What units of time? I am guessing not 2^32 seconds as that works out at about 136 years. Microseconds maybe? Or maybe I am just being thick and missed something?

  6. Henry Wertz 1 Gold badge


    Well, for computer science studies of algorithms you've got in terms of program runtime O(n) (linear), O(n^2) O(n^3) (cubic time), O(log(n)) etc. For theoretical work you don't worry about n, if there's a n^3 algorithm for doing something you try to find a way to do it that's n^2 or n instead.. So iit takes 2^32 runs through an algoirthm that takes time n. Yeah, if the algorithm takes 1000 cycles it would run 1,000,000 times a second at 1ghz, "n" would be 1 microsecond... Faster CPUs would cut the time. Pipeline stalls and inefficiencies can drop a CPU well under 1 instruction per cycle, slowing things down. Optimizations could cut the run time a lot (both the compiler variety and "hand-optimized assembly" variety), conversely if the algorithm's really complex it could take well over 1000 cycles. So 0.1-50 microseconds or so would be my guesstimate on a likely range.

    At 1 microsecond per run it'd take about an hour and 10 minutes. , it needs 64MB of data and 1GB of RAM.

This topic is closed for new posts.