Paul Clark
10/10 polish: Compare Mode, Kasiski/Friedman, rich dataset schema, expanded tests/docs
346ef1a A newer version of the Gradio SDK is available: 6.14.0
Sample Ciphertexts
Use these in Detect Mode or Compare Mode. Expected answers are at the bottom — try to call them yourself first using Explain Mode.
1 — Caesar / ROT
WKLV LV D FODVVLFDO FDHVDU FLSKHU GHPR IRU FLSKHU GHWHFWLYH DL
2 — Atbash
GSV XLWV RH ZOO BLFIH GSV VEVIVHG RMP
3 — Vigenère (short, deliberately hard)
LXFOPVEFRNHR
4 — Rail Fence
TEITELHDVLSNHDTISEIIEA
5 — Columnar transposition
EOACT IPTRH IIEEN HSGES SOSCR REMEN AERTC OEFNT TYIHE THCMC
6 — Plaintext
THE LIBRARY PRESERVES KNOWLEDGE FOR THE COMMUNITY
7 — Affine
IZZWVU NWHJUS NSV BUKUSO YUSL NWHJUSE
8 — Monoalphabetic substitution
GUF KSCQNQA HQFDFQXFD ZRMVKFTBF YMQ GUF EMSSWRSGA
Expected answers
| # | Cipher | Notes |
|---|---|---|
| 1 | Caesar shift 3 | Brute-forces in 26 tries. Chi-squared confirms. |
| 2 | Atbash | Self-inverse — a single decode test wins. |
| 3 | Vigenère, key LEMON |
Sample is too short for high confidence — that's the lesson. |
| 4 | Rail-fence (3 rails) | English letters, weak bigrams: classic transposition signature. |
| 5 | Columnar transposition | Same family as #4; bigram support stays low. |
| 6 | Plaintext | Should classify confidently as plaintext. |
| 7 | Affine, a=5, b=8 | 312-key brute force; the affine candidate table surfaces it. |
| 8 | Monoalphabetic substitution | English-like IoC but disrupted bigrams. |