Kryptos K4 · Unsolved since 1990

Research Toolkit

40 cryptanalytic strategies. 10+ cipher families eliminated. 5M+ running-key trials. The most systematic open-source assault on K4.

The Ciphertext

97 Unsolved Characters

Unknown (73 chars) EASTNORTHEAST (pos 22–34) BERLINCLOCK (pos 64–74)
40 Strategies
10+ Families Eliminated
5M+ Running-Key Trials
24 Known Characters
0.036 Index of Coincidence

Systematic Elimination

What K4 Is NOT

Each row represents a cipher family tested to destruction against the 24 known plaintext characters.

Cipher Family Status Evidence
Simple Monoalphabetic✗ EliminatedMapping contradiction at known positions for every transposition (0/13,136 consistent)
Affine Cipher✗ EliminatedNo valid (a,b) pair satisfies 24 known positions
Autokey Vigenère✗ EliminatedZero consistent configurations across all transpositions
Gromark Cipher✗ Eliminated22,000 primers (2-5 digit seeds), all fail at anchors
Hill 2×2 Cipher✗ Eliminated157,248 invertible matrices tested, zero produce ≥16 anchor matches
Single Vigenère✗ EliminatedNo single period satisfies all 4 anchor groups simultaneously
Carter Diary Running Key✗ EliminatedFull Nov 26, 1922 diary — 147K attempts, best 7/24 = noise
Carter Journals (All 3 Seasons)✗ Eliminated77K alpha chars from Griffith Institute — 5M+ attempts, best 8/24 = noise
K1-K3 Plaintexts as Key✗ EliminatedDirect running-key test, no anchor matches
Mathematical Key Derivation✗ EliminatedCoordinates, Fibonacci, primes, modular, positional — none fit forced key
Pure Playfair△ Unlikely97 chars (odd), double letters (BB, SS), best 7/9 partial match
Pure Four-square / Two-square△ UnlikelyNo anchor matches under any keyword configuration
Polyalphabetic + Transposition? PossibleConsistent configs exist but no English plaintext outside anchors
Running Key (Unknown Source)? PossibleKey appears text-derived — source not yet identified
Modified Vigenère? Possible"Make the Vigenère more abstruse" — per Scheidt
3+ Layer Compound? UntestedEvery 2-layer combo searched; third layer not yet attempted

Analysis

Key Findings

The Forced Key

BLZCDCYYGCKAZMUYKLGKORNA

Under Vigenère, the 24 known plaintext chars force these key values. No mathematical function (linear, quadratic, Fibonacci, prime, coordinate-derived) fits them. The key is almost certainly text-derived.

Period-26 Near-Miss

With key length 26, only 1 conflict exists (B vs A, off by exactly 1). 23 of 26 key slots are determined. Periods 27-29 have zero conflicts but more unknowns. This could indicate a Quagmire variant with nonstandard indexing.

K3 ≠ Carter Diary

Sanborn made 9 deliberate creative changes when adapting Carter's diary for K3: changed pronouns (we→I), substituted words (top→upper, candle→flame), added emotional phrases. These are artistic choices, not errors.

IC Below Random

K4's Index of Coincidence (0.036) is below pure random (0.038), far below English (0.067). This suggests multi-layer encryption or a very long/random key.

Arsenal

40 Strategies

Each strategy is a standalone Python script producing structured JSON output.

Discovery

Sanborn's Creative Changes to the Carter Diary

K3 PlaintextActual DiaryChange
I MADE A TINY BREACHWE MADE A TINY BREACHwe → I
IN THE UPPER LEFT HAND CORNERIN THE TOP LEFT HAND CORNERtop → upper
I INSERTED THE CANDLE AND PEERED INI WIDENED THE BREACH AND BY MEANS OF THE CANDLE LOOKED INSimplified
THE FLAME TO FLICKERTHE CANDLE TO FLICKERcandle → flame
PRESENTLY DETAILS ... EMERGED FROM THE MISTONE'S EYES BECAME ACCUSTOMED TO THE GLIMMER OF LIGHT ...Complete rewrite
SLOWLY DESPARATLY SLOWLY(not in diary)Added by Sanborn
WITH TREMBLING HANDS(not in diary)Added by Sanborn

Open Questions

What Could K4 Be?

Modified Vigenère + Transposition

The leading hypothesis. Consistent configurations exist, but no key source has been identified. Ed Scheidt confirmed methods from Friedman's Military Cryptanalysis texts.

Unknown Running Key Source

The forced key values look random — exactly what a running key from natural text produces. Carter is eliminated. Untested: Friedman's texts, Sanborn's other sculptures, CIA documents.

Three or More Layers

Every two-layer combination has been searched. A third layer (fractionation, additional transposition) could explain the below-random IC.

Sanborn's Own Invention

Something close to but not exactly standard Vigenère. The ±1 offset at period 26 could be a signature of a modified algorithm.