Call for review: Hashing by hand algorithm
in [Cryptography]
|
Prev: Dynamic Hill cipher
Next: PE Scrambler
From: J.D. on 29 Apr 2010 03:24 On Apr 29, 2:48 am, MrD <mrdemean...(a)jackpot.invalid> wrote: > J.D. wrote: > > > The sample text I decided to use for hashing is as follows (omitting > > all spaces, numbers and non-letter characters): > > DIFFERENTIALCRYPTANALYSISOFDESLIKE > > CRYPTOSYSTEMSELIBIHAMADISHAMIRTHE > > DATAENCRYPTIONSTANDARDDESISTHEBEST > > KNOWNANDMOSTWIDELYUSEDCRYPTOSYSTEM > > FORCIVILIANAPPLICATIONSITWASDEVELOPED > > ATIBMANDADOPTEDBYTHENATIONALBURAEU > > OFSTANDARDSINTHEMIDSEVENTIESANDHAS > > SUCCESSFULLYWITHSTOODALLTHEATTACKS > > PUBLISHEDSOFARINTHEOPENLITERATUREIN > > THISPAPERWEDEVELOPANEWTYPEOF > > CRYPTANALYTICATTACKWHICHCANBREAKDES > > WITHUPTOEIGHTROUNDSINAFEWMINUTESON > > APCANDCANBREAKDESWITHUPTOFIFTEEN > > ROUNDSFASTERTHANANEXHAUSTIVESEARCH > > THENEWATTACKCANBEAPPLIEDTOAVARIETY > > OFDESLIKESUBSTITUTIONPERMUTATION > > CRYPTOSYSTEMSANDDEMONSTRATESTHE > > CRUCIALROLEOFTHEUNPUBLIHEDDESIGNRULES > > (you may recognize this as the title, authors and abstract of a famous > > paper in cryptanalysis) > > "Bureau" is mis-spelt (this might conceivably affect your results). > > -- > MrD. Indeed it was misspelled (as is "unpublihed" (sic), now that I look at it). The misspelling of "buraeu" is not my fault -- it was misspelled in the paper I took the text from. Contrarily, "unpublihed" is all me. Fortunately, misspellings in the sample text (regardless of origin) could not affect the test I was conducting (which was purely to find a rough estimate of how fast the algorithms are).
From: WTShaw on 29 Apr 2010 06:32 On Apr 26, 8:04 am, bmearns <mearn...(a)gmail.com> wrote: > On Apr 25, 2:05 pm, WTShaw <lure...(a)gmail.com> wrote: > [snip] > > > > > Perhaps it is not so simple, but you probably don't want my opinion. > > Only if it's on topic. Most of it wasn't (I'm shocked!), but you did > have one useful morsel. > > > We I saw this post, I was amused because during a thunderstorm the > > other night the computers were safely down and I did not want to waste > > batteries. To confirm a counted hash, I had done it hours before by > > hand, not so easy as I have trouble writing. > > > The hash is doubly processed and the result two steps away from the > > normal alphabetic sequence is > > xjzlacokyungphfsbreiwmvdtq. > > WTShaw, it never ceases to amaze me how you never think to actually > describe any of the algorithms you're talking about. > > > I doubt that you can do much with it but the lessons are clear, error > > prone if you don't have a way to prove the results. > > Hey, you actually posted something relevant and useful! Congrats! > > You're right, hashing by hand is error prone, and that's a serious > issue. If it takes 15 minutes to complete a hash only to find out you > got it wrong, it's a pretty terrible waste of time. Worse still if you > don't know whether you got it right or wrong. As it so happens, I went > through my algorithm by hand several times and got it wrong just about > every time. And it wasn't fast. And it wasn't even with a full 52-card > deck. Alas, I'm afraid my attempt was largely fruitless except for the > always useful lessons learned. But I need to discuss that in another > post. > > [snip] > > -Brian I see that you guy have been through some difficulty working on your idea. I wholeheartedly render my sympathies. OK, back to describing the best technique for hand-working a counted hash. First, let me say that it is a very old idea and I have references to the idea in the 30's but not the name. A better description is from a the document stamped top-secret, I know because I have one. The includes the M-94 and how to shuffle the wheels based on a sentence or just a repeated phrase, the last part that I have dubbed Projection. I learned of how the hash worked about 1954 as he used it during the war. Not having a better name, I named it Counted Hash years ago and did describe it here. 0)The best method for doing a counted hash uses a large graph of maybe 30 cells on a side. 1) Write 26 characters to be hashed centered on the top, 2 blank cells cells at each end. 2) Center the abc.. series on left and right vertical columns of the paper, blank cells top and bottom. 3) In each row when the letter from the top is the same as the letter on the left or right edge, write the letter of the moment. 4) When you have filled the chart, you should have used each of the top letters once somewhere in the column below it dependent on finding the appropriate row.. 5) Beginning with "A' row, going left to right, number the letters 1-26. If you have 3 A's, they would be 1-2-3. B would start with 4 if you had any B's. Some rows would have not entries it those letters did not occur at the top. 6) At this point you might want to number the centered characters 1-26 on the left edge by writing 1-26 in the second vertical column which should have been left blank. 7) At the bottom of the page, find the number of the character assigned as you looked at each row in step 5) and write it in the next to bottom row. 8)Write the letter in the bottom row that corresponds to the the letter-number on the left edge from the number in each cell in the row above the bottom row. 9) You should have written on the bottow row the completed hash, a permutation of 26 letters. 10) To chain another hash, do as before except use the just completed hash for the vertical alphabetic sequences on the left and right edges of the new grid.
From: bmearns on 29 Apr 2010 08:55 On Apr 29, 2:48 am, MrD <mrdemean...(a)jackpot.invalid> wrote: > J.D. wrote: > > > The sample text I decided to use for hashing is as follows (omitting > > all spaces, numbers and non-letter characters): > > DIFFERENTIALCRYPTANALYSISOFDESLIKE > > CRYPTOSYSTEMSELIBIHAMADISHAMIRTHE > > DATAENCRYPTIONSTANDARDDESISTHEBEST > > KNOWNANDMOSTWIDELYUSEDCRYPTOSYSTEM > > FORCIVILIANAPPLICATIONSITWASDEVELOPED > > ATIBMANDADOPTEDBYTHENATIONALBURAEU > > OFSTANDARDSINTHEMIDSEVENTIESANDHAS > > SUCCESSFULLYWITHSTOODALLTHEATTACKS > > PUBLISHEDSOFARINTHEOPENLITERATUREIN > > THISPAPERWEDEVELOPANEWTYPEOF > > CRYPTANALYTICATTACKWHICHCANBREAKDES > > WITHUPTOEIGHTROUNDSINAFEWMINUTESON > > APCANDCANBREAKDESWITHUPTOFIFTEEN > > ROUNDSFASTERTHANANEXHAUSTIVESEARCH > > THENEWATTACKCANBEAPPLIEDTOAVARIETY > > OFDESLIKESUBSTITUTIONPERMUTATION > > CRYPTOSYSTEMSANDDEMONSTRATESTHE > > CRUCIALROLEOFTHEUNPUBLIHEDDESIGNRULES > > (you may recognize this as the title, authors and abstract of a famous > > paper in cryptanalysis) > > "Bureau" is mis-spelt (this might conceivably affect your results). > > -- > MrD. Thanks for catching that, MrD. It doesn't actually impact our results, but it's good know. Cheers, -Brian
From: bmearns on 29 Apr 2010 09:15 On Apr 29, 6:32 am, WTShaw <lure...(a)gmail.com> wrote: [snip] > > I see that you guy have been through some difficulty working on your > idea. I wholeheartedly render my sympathies. OK, back to describing > the best technique for hand-working a counted hash. First, let me say > that it is a very old idea and I have references to the idea in the > 30's but not the name. A better description is from a the document > stamped top-secret, I know because I have one. The includes the M-94 > and how to shuffle the wheels based on a sentence or just a repeated > phrase, the last part that I have dubbed Projection. I learned of how > the hash worked about 1954 as he used it during the war. Not having a > better name, I named it Counted Hash years ago and did describe it > here. > > 0)The best method for doing a counted hash uses a large graph of maybe > 30 cells on a side. > > 1) Write 26 characters to be hashed centered on the top, 2 blank cells > cells at each end. > > 2) Center the abc.. series on left and right vertical columns of the > paper, blank cells top and bottom. > > 3) In each row when the letter from the top is the same as the letter > on the left or right edge, write the letter of the moment. > > 4) When you have filled the chart, you should have used each of the > top letters once somewhere in the column below it dependent on finding > the appropriate row.. > > 5) Beginning with "A' row, going left to right, number the letters > 1-26. If you have 3 A's, they would be 1-2-3. B would start with 4 if > you had any B's. Some rows would have not entries it those letters > did not occur at the top. > > 6) At this point you might want to number the centered characters 1-26 > on the left edge by writing 1-26 in the second vertical column which > should have been left blank. > > 7) At the bottom of the page, find the number of the character > assigned as you looked at each row in step 5) and write it > in the next to bottom row. > > 8)Write the letter in the bottom row that corresponds to the the > letter-number on the left edge from the number in each cell in the row > above the bottom row. > > 9) You should have written on the bottow row the completed hash, a > permutation of 26 letters. > > 10) To chain another hash, do as before except use the just completed > hash for the vertical alphabetic sequences on the left and right edges > of the new grid. WTShaw- Your recent bouts of lucidity are a pleasant surprise. I think I followed your description, Let me try a short example (view in fixed- width font): --debbcfiha-- - a a (1) b bb (2,3) c c (4) d d (5) e e (6) f f (7) g h h (8) i i (9) - 562347891 - efbcdehia I put the numberings of each cell off to the side, but I think this is what you were describing. It's interesting, but highly biased. An a in the digest will almost always correspond to a in the message (anytime there is an a in the message). Just generally, the letters in the digest tend to be pretty close to the corresponding letter in the message. I'm not saying I could uniquely reverse it, but it seems that a significant amount of information about the message can be pretty easily and reliably gleaned from the digest. It's also extremely non-ideal that you have to write everything down. -Brian
From: J.D. on 29 Apr 2010 13:25
On Apr 29, 12:34 pm, bmearns <mearn...(a)gmail.com> wrote: > > So that's the rake operation, and the overall algorithm is simply > feeding in input cards using the old shuffle_1 routine, Which one was the old shuffle_1 routine? I've lost track of what was what... |