How to solve code puzzles?

[Highroller]

I find that everytime a puzzle requiring you to decipher a message is posted on this site I am completely baffled by it. How does one go about solving these kinds of problems? Is there a set technique or procedure for it?

[Persona]

Depends on the type of message. I generally assume that one of three encodings are in effect:

1) Caesar shift (also called rot-#). This means taking every letter in the original message and moving them forward or backward in the original message. So a Caesar shift of 1 turns ABCD into BCDE. The most common of these online is the rot-13, because the encryption and decryption are identical. (ABCD into NOPQ, NOPQ into ABCD)

2) Keyword Mixed Sequence. This still converts each plaintext letter to a single ciphertext letter, but the pattern isn't so simple. Say the key word is LUMBERJACKS. You'd write that word out, then follow it with all of the letters that aren't in LUMBERJACKS in alphabetical order:

code:

---

LUMBERJACKSDFGHINOPQTVWXYZ


abcdefghijklmnopqrstuvwxyz

---

The lower alphabet could be just the direct standard alphabet, or that alphabet reversed, or even another mixed keyword alphabet. The good news is, each letter in the plaintext always and only goes to one letter in the ciphertext.

3) Vigenere Cipher. This one seems to be popular in the IT threads. Say your plaintext is ATTACK AT DAWN. You take a key word, like ABLE, and repeatedly write it over the message:

code:

---

ABLEABLEABLE


ATTACKATDAWN

---

Then, you treat each column as an addition, where A=1, B=2...Z=26 (Rarer, but possible, is to treat A=0, B=1, ... Z=25). Then you convert the total back into the appropriate letter.

code:

---

ABLEABLEABLE


ATTACKATDAWN


------------


BVEFDMMYECHS

---

As you can see, this breaks the "one letter ->one letter" that made the first two encryption schemes so pleasant.

But enough theory. Here's the basic recipie I follow in practice. If I assume that a message is encoded with one of the first two schemes:

1) I go to One Across and feed some of the longer words into their Cryptogram engine. This may turn something up, if you're in case 1 or 2.

2) Are there letter patterns I might recognize? One recent puzzle I knew was based on recent headlines. I saw the pattern DBEFEBKO. Knowing what I know about the last month, I tried HAITIANS, and the rest of the puzzle fell into place easily. One letter (a, I) and two letter words (preopositions) also work very well for this.

3) I look at the distribution of the letters. Are there some letters that appear frequently? Those are probably one of the letters in SENORITA. Do some not appear at all? Try the high-scoring Scrabble letters.

If it's a Vigenere cipher and I have no clues about what the key might be:
1) I look for groups of letters (2 or more) that are repeated within the message. Ideally, I find a couple longer groups, that are all some multiple of the same distance away. This would give me a clue as to the length of the key. So if I found a 6-long repeat with a 45 character gap, a 4 long repeat at a 20 character gap, and a few 3 letter repeats at 30, 55, and 25, I'd be prety confident that the key word was only 5 letters long.

2) Once I know the length of the key, I write the message out with that many characters to a line, so that all the characters in that line are affected by a single letter. Then, each column is just a different caesar shift, although the plaintext is every nth word of the original message. This is obviously trickier than a simple shift, but with the frequency counts and trying to chain words across columns, it can work.

At least, that's how I do it.

[jesternl]

Thanks for a very clear explanation!

[Highroller]

You mentioned a cryptogram engine. Do you know where I can find a useful one?

[Lucky Wizard]

The one he mentioned is at http://www.oneacross.com/cryptograms/search.html , but I'm sure there are others. You can probably find another one by going through http://directory.google.com/Top/Games/Puzzles/Cryptograms/?tc=1/ .

BTW, another cipher that has been used at least twice in the IT threads is the Playfair cipher. http://www.trumpetpower.com/Papers/Crypto/Playfair

I don't know any strategies for deciphering Playfair ciphers.

By the way, a book that might interest and help you is Cryptography by Laurence Dwight Smith. It certainly got me interested in these ciphers. Although the book is old, it's still useful.

Just for the record, I usually see the Vigenere cipher used with A=0, B=1, etc.

[Lepton]

TLC, I have snipped your post off of this topic. It is now a thread of its own in the Visitor Submitted Puzzles forum: link.