This technique was invented by Mauborgne and Vernam in 1917. The basic idea is that you have a huge pool of truly random non-repeating key letters typed onto sheets of paper and glued up into pads. There are two copies and only two copies of each pad. The sender has one and the receiver has the other. Each key letter on the pad is used by the sender to encrypt one and only one plaintext letter. Encryption comprises giving each letter a number (A = 1 to Z = 26), adding the plaintext characters number to the key characters number, subtracting 26 if necessary to bring the number back into the 1 to 26 range, and turning it back into a letter. For example, Plaintext T = 20, Key N = 14, so Ciphertext is the letter corresponding to 34. Since there isnt a letter number 34, we subtract 26, to give 8. H is the 8th letter of the alphabet, so the Cyphertext is H. Now move on to the next letter in the plaintext and the next letter on the pad. When the message is complete, all sheets of the pad used in its encryption are destroyed. The message can now safely be transmitted by whatever means are available and appropriate. The receiver takes the ciphertext H, turns it into a number (8), consults his copy of the one-time pad to obtain the key for that character, and subtracts it from 8. 8 - 14 is -6. No such letter, so add 26 to bring it into range. This yields 20. Behold, T!
Because each letter is encrypted using a different key, there is no way to reconstruct the message from the ciphertext. Or rather, all messages of that length can be reconstructed - and for a ciphertext of any length there are millions of trillions of messages and no way to tell which is the correct one. Yes, you could generate the list of messages of that length and try to guess which it is. You could do that without having the ciphertext at all! It doesnt help the cryptanalyst one jot. The ciphertext
could equally mean
or any other message of the same length, including a shopping list or a love letter.