Cipher Decoder

// You have the message, not the method

Most cipher tools start by asking a question you cannot answer. Which cipher is it? What is the key? If you knew that, you would be most of the way there already. What you usually have is a string of letters from a puzzle box, a geocache, a CTF, or a friend who thinks they are funny, and no idea what was done to it.

This page starts from the other end. Paste the message and it runs every cipher it knows, scores each result for how much it reads like English, and shows you the ones worth looking at, best first. Each result says which cipher produced it and what key or shift it needed, so you learn the method along with the message. If you already know the cipher, the manual tab runs any of the fourteen directly, in either direction.

// How a computer decides that something reads like English

Two measurements, combined. The first is letter frequency. English is lopsided: E is about 12 percent of letters, T about 9, while Z and Q are under a tenth of a percent each. A chi-squared test measures how far a candidate's letter distribution sits from that shape. Real English usually lands between 10 and 30; a random jumble lands between 80 and 200. That single number is enough to pick the right Caesar shift out of 25 almost every time.

Frequency alone can be fooled, though, because rearranging letters does not change how often each one appears. So the second measurement counts how many words in the result are among the hundred commonest English words. A genuine decode is full of the, and, of, to. An anagram of one is not. The score you see is 60 percent frequency and 40 percent real words, which is why a transposition cipher like a rail fence gets caught rather than sailing through.

Vigenere needs a third idea, because it uses a different shift for every letter of its key and flattens the frequency curve that the first test relies on. The index of coincidence measures how often two letters picked at random from a text match: about 0.067 for English, about 0.038 for random letters. Split the ciphertext into every second letter, every third, every fourth, and so on, and at the true key length each group is a plain Caesar shift, so its index jumps back toward English. That reveals the length; each group's own best shift then gives one letter of the key.

// What it can read, and what it cannot

Fourteen in total. Classical substitution and transposition: Caesar, ROT13, ROT47, Atbash, Vigenere, Affine, and rail fence. Codes: Morse, A1Z26, and Bacon's cipher. Encodings: Base64, Base32, hexadecimal, and binary. The encodings are recognised by their alphabet alone, so a run of dots and dashes or a string of ones and zeroes is identified before any letter frequency is measured.

The honest limits are worth stating. Vigenere key recovery needs a reasonable amount of text, at least forty letters, and it is refused below that rather than guessed at, because a key fitted to a dozen letters is noise wearing a costume. Anything using a keyword substitution alphabet, a one-time pad, or a modern algorithm is not going to fall to frequency analysis and is not attempted. And when nothing scores well enough, the page says it has no confident decode instead of presenting its least-bad guess as an answer.

// fair questions

How do you decode a Caesar cipher without knowing the key?

You try all of them. There are only 25 useful shifts, so a computer can produce every possibility instantly and rank them by how much each reads like English. The full table of 25 is on this page, with the best-scoring shift marked, so you can confirm the answer with your own eyes rather than taking the ranking on trust.

What is the difference between ROT13 and a Caesar cipher?

ROT13 is a Caesar cipher with the shift fixed at 13. Because 13 is exactly half of 26, applying it twice returns the original text, which makes the same operation both the encoder and the decoder. That self-inverse property is why it became the standard way to hide spoilers and punchlines in plain text.

Can a Vigenere cipher actually be cracked?

Yes, given enough text, and this page does it without being told the key. It uses the index of coincidence to work out the key length, then solves each key position as its own Caesar shift. It needs at least forty letters and works better with a few hundred. Below that threshold it declines rather than inventing a key, because a key fitted to a handful of letters is fitted to noise.

How do I know which cipher was used?

Often the character set gives it away before anything else: dots and dashes mean Morse, only ones and zeroes means binary, A to F and digits means hexadecimal, and trailing equals signs suggest Base64. Past that it is statistical. Auto-solve runs all fourteen and reports which produced the most English-looking result, along with the key or shift it needed, which is usually the fastest way to identify an unknown cipher.

Is it safe to paste puzzle or CTF text here?

Yes. Every cipher and the whole detection pipeline run in your browser on your own machine. Nothing you paste is sent to a server, logged, or stored, so the text never leaves the page you are looking at.

Can I encode a message as well as decode one?

Yes. Switch to the manual tab, pick a cipher, and choose encode. Every cipher here runs both directions, with the parameters you set, so you can produce a Caesar message with a specific shift or a Vigenere message with your own key.