The ancient world¶

From about 1900 BC to the 700s AD: scribes who bent their scripts, Greeks who hid messages under wax and hair, a Spartan rod, a Greek signal grid, Caesar's shift of three, and secret scripts in India, Persia and Egypt.

People in the ancient world kept secrets in three ways, and all three are still in use:

Hide the message. If nobody knows a message exists, nobody tries to read it. This is steganography, from the Greek for "covered writing".
Replace the words. Agree in advance that a sign or phrase stands for a whole meaning: "the lamp is lit" means "attack at dawn". This is a code.
Scramble the letters. Swap each letter for another (substitution) or keep the letters but shuffle their order (transposition). This is a cipher.

The readable message is the plaintext, the scrambled one the ciphertext, and the small secret that says how to scramble it (which rod, which shift) is the key. Nearly everything on this page is weak by modern standards. What matters is that each of these ideas appears here first.

When	Where	What	Kind
c. 1900 BC	Beni Hasan, Egypt	odd hieroglyphs in the tomb of Khnumhotep II	changed script, not secret
1600s–1200s BC (disputed)	Babylonia	a glaze or glass recipe in rare sign values	trade secret
600s BC on	Sparta	the skytale message staff	transposition (disputed)
600s–500s BC or later	Judah	Atbash names in Jeremiah	substitution
c. 499 BC	Persia to Miletus	Histiaeus's tattooed slave	steganography
480s BC	Persia to Sparta	Demaratus's blank wax tablet	steganography
350s BC	Greece	Aeneas Tacticus's chapter on secret messages	steganography, codes
100s BC	Greece	Polybius's torch-signal square	letters as numbers
50s BC	Rome	Caesar's shift of three	substitution
400s–200s BC	China	secret tallies and split letters	codes
c. AD 300	India	secret writing among the Kama Sutra's 64 arts	substitution
AD 224–651 (reported later)	Sasanian Persia	a secret script for the kings' letters	secret alphabet
300s–700s	Egypt	Coptic and Greek letter-swap and number ciphers	substitution

Egypt: the scribe who changed the signs¶

Khnumhotep II, about 1900 BC¶

David Kahn begins the historical part of The Codebreakers, still the standard history of the subject, in a tomb at Beni Hasan on the Nile, near the old town of Menet Khufu. Khnumhotep II was a regional governor under the 12th Dynasty pharaohs Amenemhat II and Senusret II, around 1900 BC. The scribe who laid out the long inscription of his life swapped some ordinary hieroglyphs for rarer, unexpected forms.

Nobody was meant to be kept out. The text sits on a tomb wall for anyone to read, and the odd signs most likely added dignity and a touch of learnedness. Kahn's point is that it is the first known deliberate change to a script, one of the two ingredients of cryptography. The other, secrecy, came later.

Enigmatic writing¶

In the New Kingdom (c. 1550–1070 BC), Egyptian priests went much further. Funerary books in the tombs of Ramesses VI and Ramesses IX, and on one of the gilded shrines that enclosed Tutankhamun's sarcophagus, are written in enigmatic hieroglyphs: familiar signs used for unusual sounds, and new signs invented for the occasion. The Egyptologist John Darnell has argued that these texts were meant to be puzzled out, which made reading them a kind of religious act, not just a matter of keeping outsiders away.

The most extreme examples are from the Roman period, when Egypt was a Roman province. In the temple of the ram-headed god Khnum at Esna, two hymns stand side by side. One is written almost entirely with ram signs and the other almost entirely with crocodile signs, both animals sacred to Khnum. Serge Sauneron published them, and Christian Leitz offered a reading in 2001. The ram hymn can now be translated reasonably well; the damaged crocodile hymn is still only partly understood.

These are not ciphers in the modern sense. There is no key, and a trained priest could read them. But they show a script being bent on purpose so that only the initiated can follow it, which is close to the heart of the idea.

Mesopotamia: a potter's recipe¶

The oldest known writing hidden for a practical reason is a small clay tablet in the British Museum (BM 120960). It was bought from a dealer and is said to come from Tell Umar, the site of Seleucia on the Tigris, in modern Iraq. C. J. Gadd and Reginald Campbell Thompson published it in 1936 as "A Middle-Babylonian Chemical Text" and read it as recipes for glazing pottery, the earliest such formulas known. The British Museum now follows a later reading, by Leo Oppenheim and colleagues in 1970, as a recipe for red glass.

Kahn describes how the writer chose uncommon readings of the cuneiform signs and spelled words in unusual ways, so that the recipe would mean little to anyone outside the trade. The motive looks familiar: a craftsman guarding a commercial secret.

Its date is uncertain. Gadd and Thompson linked a royal name on the tablet to King Gulkishar and dated it no later than the 17th century BC; Kahn gives about 1500 BC; the British Museum's catalogue says 1400–1200 BC.

In the last centuries of cuneiform, around the Seleucid period (after 312 BC), scribes also wrote some signs as numbers. Tablets survive that pair each number with a syllable, a ready-made table for a number cipher. Laurie Pearce studied this practice in her 1982 Yale thesis, Cuneiform Cryptography: Numerical Substitutions for Syllabic and Logographic Signs.

The Hebrew Bible: Atbash¶

Swapping the alphabet end to end¶

Atbash swaps the first letter of the Hebrew alphabet with the last, the second with the second-to-last, and so on. The name is its own example: alef ↔ tav, bet ↔ shin. Applying it twice gives back the original, so the same table encrypts and decrypts.

The Book of Jeremiah uses it at least twice:

Sheshach (ששך) in Jeremiah 25:26 and 51:41 is Atbash for Babel (בבל), Babylon.
Leb-kamai (לב קמי) in Jeremiah 51:1 is Atbash for Kasdim (כשדים), the Chaldeans.

Ancient readers knew this. The Aramaic Targum puts "Babel" where the Hebrew has Sheshach, and the Greek Septuagint translates Leb-kamai simply as "Chaldeans". The swap can't have been for secrecy: Jeremiah 51:41 names Babylon openly in the same verse. It reads more like wordplay, a way of saying "Babylon" with a sneer, and some scholars think the forms were added by later copyists. The book's core goes back to the late 7th and early 6th centuries BC, but its exact editing history is debated.

Two relatives use the same idea with different pairings. Albam splits the alphabet in half and pairs alef with lamed, the first letter of the second half. Atbah uses yet another fixed pairing. All three are monoalphabetic: each letter always becomes the same other letter.

This little program does Atbash on the Latin alphabet, then on Hebrew. Hebrew has five letters with a different shape at the end of a word, so the code turns those back into normal letters first and gives the last letter its final form afterwards.

latin = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
hebrew = "אבגדהוזחטיכלמנסעפצקרשת"        # the 22 letters, alef to tav
to_plain = {"ך": "כ", "ם": "מ", "ן": "נ", "ף": "פ", "ץ": "צ"}
to_final = {v: k for k, v in to_plain.items()}

def atbash(word, alphabet):
    word = "".join(to_plain.get(ch, ch) for ch in word)   # final forms count as normal letters
    out = "".join(alphabet[-1 - alphabet.index(ch)] for ch in word)
    return out[:-1] + to_final.get(out[-1], out[-1])     # a Hebrew word ends in a final form

print(atbash("SECRET", latin), atbash(atbash("SECRET", latin), latin))
print("Babel", "בבל", "becomes", atbash("בבל", hebrew), "Sheshach")
print("Kasdim", "כשדים", "becomes", atbash("כשדים", hebrew), "Leb-kamai")

Output:

HVXIVG SECRET
Babel בבל becomes ששך Sheshach
Kasdim כשדים becomes לבקמי Leb-kamai

The Dead Sea Scrolls¶

Centuries later, the community at Qumran, by the Dead Sea, wrote some of its scrolls, probably in the 2nd or 1st century BC, in a substitute alphabet scholars call Cryptic A. Each Hebrew letter has its own invented sign. Scholars learned to read it in the 1950s. How many scrolls use it is disputed: Stephen Pfann's 2000 edition counted 36 manuscripts, most of them tiny fragments, while Jonathan Ben-Dov argued in 2024 that only eight or nine are really in the script. In 2017 Eshbal Ratzon and Ben-Dov pieced one of them, a calendar, back together from 42 tiny fragments. Nothing in the cryptic scrolls is especially secret; the script seems to have marked texts meant for the community's teachers, a badge of rank more than a lock.

Greece: hidden messages¶

Herodotus's stories¶

Herodotus, writing around 430 BC, tells the oldest stories of hidden messages. They are good stories, and he heard most of them second-hand, so treat the details with care.

The tattooed slave (Histories 5.35). Histiaeus, the former ruler of Miletus, was being kept at the Persian court in Susa. Around 499 BC he wanted to tell his son-in-law Aristagoras, now running Miletus, to revolt against Persia. He shaved the head of his most trusted slave, marked the message on his scalp, waited for the hair to grow back, and sent him with one instruction: tell Aristagoras to shave my head. The Ionian Revolt followed.
The blank tablet (7.239). Demaratus, an exiled king of Sparta living in Persia, learned that Xerxes was planning to invade Greece. He took a folding wooden writing tablet, scraped off the wax, wrote his warning on the bare wood, and covered it with fresh wax. The blank tablet passed the guards. In Sparta nobody could work out what it meant until Gorgo, the wife of King Leonidas, suggested scraping off the wax. Herodotus himself wonders whether Demaratus sent it out of goodwill or to gloat, and some editors have suspected the whole chapter is a later addition.
The hare (1.123). Harpagus, a Median noble, sewed a letter into the belly of a freshly killed hare and sent it to Cyrus in Persia with a servant dressed as a hunter.
The arrows (8.128). During the Persian siege of Potidaea in 480–479 BC, the traitor Timoxenus and the Persian general Artabazus wrapped letters around the notched ends of arrows and shot them to an agreed spot. The plot came out when one arrow missed and hit a man in the shoulder.

Every one of these methods fails the moment someone knows where to look: shave the head, scrape the wax, cut open the hare. That weakness never goes away. Steganography is only as safe as the trick is unknown.

Aeneas Tacticus: a chapter on secret messages¶

The first known writer to collect methods on purpose is Aeneas, called Tacticus, who wrote a manual around the 350s BC on how a city could survive a siege. Chapter 31 is about sending messages unseen. Among his ideas:

prick tiny holes above or below letters in a book or an ordinary letter; the receiver copies out the marked letters in order;
write on a leaf and bind it under a bandage on a wound;
write on thin sheets of lead and wear them as earrings;
sew a letter into the sole of a sandal;
write on a bladder, then inflate it inside an oil flask so it presses against the walls and looks like part of the flask;
bore 24 holes in a knucklebone (an astragal) for the 24 letters and pass a thread through them in the order of the message. The receiver pulls the thread out and reads the letters off the holes.

The pricked-letter trick had a long life. In 19th-century England, where a newspaper went through the post far more cheaply than a letter, people are said to have pricked holes over letters in newspapers to send messages for the price of the paper.

Ovid, the Roman poet, gives lovers the invisible-ink version in the Art of Love (Book 3): write in fresh milk, and the reader makes it appear with coal dust.

Aeneas's water clocks: a code book in a jar¶

Polybius, in the 2nd century BC, describes a signal system he credits to Aeneas (Histories 10.44). Each side had an identical jar of water with a cork float carrying an upright rod. The rod was marked in sections, each with an agreed message: "cavalry have entered the country", "heavy infantry", "ships", "grain". The sender raised a torch, both sides pulled their plugs at once, and when the water had fallen to the right section the sender raised a torch again and both closed their jars. The section level with the jar's rim was the message.

This is a code: a short list of agreed meanings. Polybius's complaint is the classic complaint about codes. It can only send what someone thought to write on the rod in advance.

Greece: the skytale¶

What Plutarch says¶

The Spartan skytale (Latin scytale) is the most famous ancient cipher device, and the most doubtful. The fullest description comes from Plutarch, writing around AD 100, in his Life of Lysander (19). When the Spartan authorities sent a general abroad, they made two wooden rods of exactly the same length and thickness, kept one, and gave him the other. To send an order they wound a long, narrow strip of parchment around their rod in a tight spiral, wrote the message along the length of the rod, and unwound the strip. Off the rod, the letters were a jumble. The general wound the strip around his own rod and read it. The Roman writer Aulus Gellius, in the 2nd century AD, describes the same thing (Attic Nights 17.9).

In modern terms this is a transposition cipher. Every letter stays itself; only the order changes. The key is the thickness of the rod, which sets how many letters fit around one turn.

Was it a cipher at all?¶

The word is much older than these descriptions. The poet Archilochus, in the 7th century BC, already calls a bad piece of news "a grieving skytale", and Thucydides (1.131) tells how the Spartans recalled their general Pausanias around 470 BC by sending a herald with a skytale. Neither says anything about winding a strip or hiding the text.

In 1998 the classicist Thomas Kelly argued in Cryptologia, in a paper titled "The Myth of the Skytale", that the early sources treat the skytale as nothing more than an official dispatch, and that the cipher story only appears in writers living hundreds of years later. Some historians think the staff served to prove that a message really came from the Spartan authorities, which is authentication rather than secrecy. Others still accept the cipher reading. The safe summary: the skytale was real, and its use as a transposition cipher is possible but not proved.

Trying it¶

Either way, it would have been a weak cipher. An enemy who knew the method only had to try rods of different thickness, or, on paper, try every row length.

def wrap(text, around):
    # Write along the rod, one row per side, then unwind: each turn of the
    # strip carries one letter from every row.
    row = -(-len(text) // around)               # letters per row, rounded up
    text = text.ljust(around * row, "X")         # pad the last row
    return "".join(text[i::row] for i in range(row))

def unwrap(strip, around):
    return "".join(strip[i::around] for i in range(around))

message = "SENDMOREMENTOTHEPASS"
strip = wrap(message, 4)
print(strip)
print(unwrap(strip, 4))

# Breaking it: try rods of every thickness
for around in range(2, 7):
    print(around, unwrap(strip, around))

Output:

SONEERTPNEOADMTSMEHS
SENDMOREMENTOTHEPASS
2 SNETNODTMHOERPEAMSES
3 SETEDSHOEPOMMSNRNATE
4 SENDMOREMENTOTHEPASS
5 SROSOTAMNPDEENMHEETS
6 STDHOPMSNNTEESEOMRAE

Only one row length gives Greek, or here English. Transposition did not die with Sparta: shuffling the order of bits is one of the two basic moves inside every modern block cipher, including AES.

Greece: Polybius and his square¶

Torches on a wall¶

Polybius, a Greek statesman who lived as a hostage and later a friend of the Romans (c. 200–118 BC), complained that fire signals could only say what had been agreed in advance. In Book 10 of his Histories (10.45–47) he describes a better system, which he credits to two men named Cleoxenus and Democleitus and says he improved himself.

The 24 letters of the Greek alphabet are split into five tablets of five letters (the last has only four). The signaller stands behind a wall with torches. First he raises torches on the left, one to five, to say which tablet. Then he raises torches on the right to say which letter on that tablet. The receiver watches through an instrument with two sighting tubes, one for each side, so he can tell left from right, and writes the letters down. Every letter becomes a pair of small numbers.

It was built for speed and distance, not secrecy. Anyone who knew the tablets could read the torches. But the idea of turning a letter into a pair of coordinates turns out to be one of the most reused ideas in cryptography. Russian political prisoners in the 19th century knocked out the row and column of each letter on their cell walls. American prisoners of war in Hanoi used the same tap code from 1965, on a \(5 \times 5\) Latin grid with C standing in for K. And in 1918 the German army's ADFGVX cipher started by turning each letter into grid coordinates, then shuffled them.

greek = "ΑΒΓΔΕΖΗΘΙΚΛΜΝΞΟΠΡΣΤΥΦΧΨΩ"     # Polybius's 24 letters
square = [greek[i:i + 5] for i in range(0, 24, 5)]
for n, tablet in enumerate(square, 1):
    print(n, " ".join(tablet))

def torches(word):
    out = []
    for ch in word:
        i = greek.index(ch)
        out.append((i // 5 + 1, i % 5 + 1))     # (which tablet, which letter)
    return out

signal = torches("ΠΟΛΥΒΙΟΣ")
print(signal)
for left, right in signal[:3]:
    print("left", "*" * left, "  right", "*" * right)
print("".join(square[t - 1][p - 1] for t, p in signal))

Output:

1 Α Β Γ Δ Ε
2 Ζ Η Θ Ι Κ
3 Λ Μ Ν Ξ Ο
4 Π Ρ Σ Τ Υ
5 Φ Χ Ψ Ω
[(4, 1), (3, 5), (3, 1), (4, 5), (1, 2), (2, 4), (3, 5), (4, 3)]
left ****   right *
left ***   right *****
left ***   right *
ΠΟΛΥΒΙΟΣ

Rome: Caesar and Augustus¶

The shift of three¶

The cipher everyone knows comes from the Roman biographer Suetonius, writing around AD 120. In his Life of Julius Caesar (56) he says that when Caesar had something confidential to say in his letters, he wrote it so that the order of the letters was changed and no word could be made out: to read it, substitute the fourth letter of the alphabet, D, for A, and so on with the others. In other words, every letter moves three places down the alphabet. Cassius Dio, writing a century later, describes the same method in his account of the war in Gaul (Roman History 40.9).

Aulus Gellius adds (Attic Nights 17.9) that Caesar wrote to his agents Oppius and Balbus in letters that looked like nonsense, and that the grammarian Probus wrote a whole commentary on the secret letters in Caesar's correspondence. That commentary is lost, which is a pity: it would tell us how Caesar's ciphers worked, from someone who had studied them.

Suetonius also records the habit of Caesar's heir. Augustus, he says (Life of Augustus 88), wrote B for A, C for B and so on, a shift of one, but instead of wrapping around he wrote AA for X, the last letter of the native Latin alphabet (Y and Z were extras, borrowed for Greek words).

Greek letters on a spear¶

Caesar's own account of the Gallic War shows a different trick. In 54 BC his legate Quintus Cicero (the orator's brother) was besieged by the Nervii. Caesar sent a Gallic horseman with a letter "written in Greek letters" so that the enemy could not read it if they caught him (Gallic War 5.48). The rider could not get through, so he tied the letter to a throwing spear and hurled it over the wall. It stuck in a tower and nobody noticed it for two days. The Latin can mean Greek script or Greek language; Dio understood it as Greek.

How strong is a shift?¶

A shift cipher has as many keys as the alphabet has letters, minus one. The Latin alphabet of Caesar's day had about 21 to 23 letters (no J, U or W), so a patient enemy could try every key in a few minutes. The program below uses the modern 26-letter alphabet and then simply tries all 25 shifts.

A = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"

def caesar(text, shift):
    out = ""
    for ch in text:
        if ch in A:
            out += A[(A.index(ch) + shift) % 26]   # move along, wrap Z to A
        else:
            out += ch                              # leave spaces alone
    return out

secret = caesar("VENI VIDI VICI", 3)
print(secret)
print(caesar(secret, -3))

# Breaking it: there are only 25 shifts, so try them all
for shift in range(1, 26):
    print(shift, caesar(secret, -shift))

Output:

YHQL YLGL YLFL
VENI VIDI VICI
1 XGPK XKFK XKEK
2 WFOJ WJEJ WJDJ
3 VENI VIDI VICI
4 UDMH UHCH UHBH
5 TCLG TGBG TGAG
6 SBKF SFAF SFZF
7 RAJE REZE REYE
8 QZID QDYD QDXD
9 PYHC PCXC PCWC
10 OXGB OBWB OBVB
11 NWFA NAVA NAUA
12 MVEZ MZUZ MZTZ
13 LUDY LYTY LYSY
14 KTCX KXSX KXRX
15 JSBW JWRW JWQW
16 IRAV IVQV IVPV
17 HQZU HUPU HUOU
18 GPYT GTOT GTNT
19 FOXS FSNS FSMS
20 ENWR ERMR ERLR
21 DMVQ DQLQ DQKQ
22 CLUP CPKP CPJP
23 BKTO BOJO BOIO
24 AJSN ANIN ANHN
25 ZIRM ZMHM ZMGM

Line 3 is the only one in Latin. This attack, trying every key, is called brute force, and it is still the first question asked of any cipher: how many keys are there? A modern 128-bit key has about \(3.4 \times 10^{38}\) of them, which is why nobody brute-forces AES. The FAQ puts numbers on how far out of reach that is.

There is no record that anyone broke Caesar's letters. That says less about the cipher than about his enemies. Few of them could read Latin, and nobody yet had a method for attacking ciphers in general.

Watchwords in the camp¶

Polybius also describes how a Roman army camp passed its nightly watchword (Histories 6.34). At sunset a soldier from the tenth maniple (company) of each class of infantry and cavalry collected a wooden tablet (tessera) with the word on it from the tribune. It was passed from maniple to maniple, and the first maniples brought it back to the tribunes before dark. The tribunes checked that every tablet came back, so if one went missing they could trace exactly where. It is a password system with a delivery receipt, an early cousin of the authentication problems that network protocols still solve.

China: tallies and split letters¶

Chinese writing has thousands of characters, which makes letter ciphers awkward. Chinese military writers relied on codes and on dividing secrets instead.

The Six Secret Teachings (Liu Tao), a military text traditionally credited to the Zhou-dynasty strategist Jiang Taigong but probably put together in the Warring States period (c. 400–220 BC), has two chapters on the subject:

Secret tallies. The ruler and his general shared a set of eight tallies of different lengths. Each length meant a fixed message, from the longest (a great victory) down to the shortest (a defeat with heavy losses). Only the ruler and the general knew the meanings; a messenger who let one leak was to be executed.
Secret letters. For anything a tally couldn't say, a letter was cut into three parts and sent by three different messengers. No one messenger could read it, and only the receiver could put it together.

Much later, the Wujing Zongyao, a military encyclopedia compiled for the Song court between 1040 and 1044, describes a code built on a poem. The two sides agreed on a 40-character poem with no repeated characters and a list of 40 standard messages, one for each character. To send a message, an officer marked the matching character of the poem with his seal.

India: secret writing as an art¶

The Kama Sutra, compiled by Vatsyayana around the 3rd century AD, lists 64 arts a cultivated person should know. One is mlecchita-vikalpa, the art of secret writing, recommended so that lovers could keep their correspondence private. A 13th-century commentary by Yashodhara describes two methods. In kautiliya, letters are swapped according to the way they sound, using the order of the Sanskrit alphabet. Muladeviya swaps a smaller set of letters in fixed pairs. Both are simple substitutions.

Kahn also notes that the Arthashastra, the old Indian manual of statecraft attributed to Kautilya, describes a spy service whose agents received their instructions in secret writing.

Persia: the kings' secret script¶

Arabic writers of the 10th century, drawing on older Persian sources, describe the scripts of the Sasanian kings (AD 224–651). Ibn al-Nadim's Fihrist (987) counts seven Persian scripts, and one of them, raz ("secret"), was by these accounts the script the kings used for their secret letters to one another. Hamza al-Isfahani gives a similar description. That is about all that is known. The names come almost entirely from these later Arabic sources, and the Encyclopaedia Iranica warns that their descriptions of the scripts are vague and often imaginative, so it is hard to say what the secret script really was.

Egypt after the pharaohs: Coptic ciphers¶

In late antique and early Islamic Egypt, roughly the 300s to the 700s, Christian scribes writing Coptic and Greek used several small ciphers. Alain Delattre's survey of them centres on one called αθβη ("athbē"), a fixed swapping of letters that, like Atbash, is named after its first letters. Scribes also used codes built on the number values of letters, Caesar-style shifts, and deliberately distorted letter shapes. They turn up in manuscripts and on ostraca, the potsherds used as scrap paper. One ostracon from Djeme, near Thebes, dated between about 500 and 750, lists cipher alphabets, and Delattre has published another that combines two ciphers in one system. Like the European monks' tricks of a few centuries later, these hid a name or a note from a casual reader, not from a determined one.

Can these be broken today?¶

Yes, every one, most of them by hand in minutes. Here is each method on this page, with the attack in plain words. The skytale and Caesar programs above show two of the attacks running.

Method	Broken today?	The attack in plain words
Egyptian enigmatic writing	not a cipher	No key: a reader who knows the sign games reads it. The Esna crocodile hymn is still only partly read, because the stone is damaged, not because of a secret.
The potter's recipe, and number signs for syllables	at once	Unusual spellings and a fixed number table. Anyone who knows the script, or has the table, reads them.
Coptic letter swaps, shifts and number codes	instantly	Each is one fixed table. Try the few known ones, or count letters as for any substitution.
Atbash and the Kama Sutra swaps	instantly	No key at all: the swap is the method. Know it and you read it.
Hidden messages (shaved head, wax tablet, hare, arrows)	as soon as you look	Nothing is scrambled. Search the courier, scrape the wax, cut open the hare. Modern versions hide data in pictures or files, and statistical tests can often spot it.
Aeneas's pricked letters	as soon as you look	Hold the page to the light and copy out the marked letters.
Code by agreed signs (water clocks, tallies, the Song poem code)	only with captured examples	Without the list, a mark means nothing. But every use is the same sign for the same message, so an enemy who sees signs and then sees what happens learns the code.
Three-part letters	by capturing all three	Each messenger carries a fragment. An enemy needs every piece: an early form of secret splitting.
Skytale	easily	Try every row length; only one gives words. Only a handful of rods are possible.
Polybius square	instantly	It turns letters into number pairs but keeps every letter's count, so it falls to the same counting as a substitution, or to no attack at all if the square is the usual one.
Caesar's and Augustus's shifts	instantly	Try all 25 shifts and keep the one in the language.

Why weak ciphers worked¶

From a modern view, everything on this page is easy to break. Caesar's shift has a couple of dozen keys, the skytale a handful, Atbash and Polybius none at all. They worked anyway, for reasons that had little to do with mathematics:

Few people could read. A captured messenger carrying a jumble of letters in a foreign script was not much of a risk.
Nobody had a theory of breaking ciphers. An enemy could guess, or torture the courier, but there was no general method for attacking a substitution. That method, counting how often each letter appears, was first written down by al-Kindi, in 9th-century Baghdad.
Secrecy of method did most of the work. If the enemy didn't know you used a rod, the rod's thickness hardly mattered. Modern cryptography rejects this idea completely: the method is assumed public and only the key is secret. See the FAQ on keeping the algorithm secret.

The ancient world left three ideas that every later page builds on: hiding a message, replacing whole meanings with agreed signs, and scrambling letters under a small secret key.

References¶

David Kahn, The Codebreakers: The Comprehensive History of Secret Communication from Ancient Times to the Internet (rev. ed., Scribner, 1996), chapter 2: Khnumhotep II, the glaze tablet, Atbash, the Greek and Roman material, India
C. J. Gadd and R. Campbell Thompson, "A Middle-Babylonian Chemical Text", Iraq 3 (1936), pp. 87–96 / A. Leo Oppenheim and others, Glass and Glassmaking in Ancient Mesopotamia (Corning Museum of Glass, 1970) / Laurie E. Pearce, Cuneiform Cryptography: Numerical Substitutions for Syllabic and Logographic Signs (PhD thesis, Yale, 1982), and "Cuneiform Number-Syllabaries", Iraq (1983)
John C. Darnell, The Enigmatic Netherworld Books of the Solar-Osirian Unity (Orbis Biblicus et Orientalis 198, 2004) / Serge Sauneron, Esna (IFAO, 1959–2009) / Christian Leitz, "Die beiden kryptographischen Inschriften aus Esna mit den Widdern und Krokodilen", Studien zur Altägyptischen Kultur 29 (2001)
Jeremiah 25:26, 51:1, 51:41 (Masoretic text, the Septuagint and Targum Jonathan) / Eshbal Ratzon and Jonathan Ben-Dov, "A Newly Reconstructed Calendrical Scroll from Qumran in Cryptic Script", Journal of Biblical Literature 136, no. 4 (2017), pp. 905–936 / Stephen J. Pfann, "Cryptic Texts", in Discoveries in the Judaean Desert 36 (Oxford, 2000)
Herodotus, Histories 1.123, 5.35, 7.239, 8.128 / Aeneas Tacticus, How to Survive Under Siege 31 / Polybius, Histories 6.34, 10.44–47 / Ovid, Ars Amatoria 3
Plutarch, Life of Lysander 19 / Aulus Gellius, Attic Nights 17.9 / Thucydides 1.131 / Archilochus, fragment 185 (West) / Thomas Kelly, "The Myth of the Skytale", Cryptologia 22, no. 3 (1998), pp. 244–260
Suetonius, Life of Julius Caesar 56 and Life of Augustus 88 / Caesar, Gallic War 5.48 / Cassius Dio, Roman History 40.9
Ralph D. Sawyer (trans.), The Six Secret Teachings on the Way of Strategy (Shambhala, 1997), "Secret Tallies" and "Secret Letters" / Wujing Zongyao (1044)
Vatsyayana, Kama Sutra 1.3 (the 64 arts), with Yashodhara's Jayamangala commentary
Alain Delattre, on cryptography in late antique Egypt, in S. Holvoet and R. Veymiers (eds.), Les écritures orientales, inventées, cryptées, détournées, oubliées, redécouvertes... (Festschrift for Christian Cannuyer, 2023), pp. 185–204 (from the library abstract) / Duke Collaboratory for Classics Computing, Digital Corpus of Literary Papyri, record 128462 (the Djeme ostracon of cipher alphabets)

Weaker sources: search summaries of the Encyclopaedia Iranica entries on the Fihrist's Persian material and on Sasanian scripts (dabire), for the raz script, Ibn al-Nadim's count of seven scripts and Hamza al-Isfahani; search summaries of the British Museum and Science Museum catalogue entries for BM 120960 (date and findspot); the Esna hymns' dating and present state of decipherment from online commentaries (attalus.org and the Esna text editions on bookdown.org), not from Sauneron or Leitz directly; Wolfgang Behr's conference slides on the Wujing Zongyao poem code; G. C. Macaulay's translation note on doubts about Herodotus 7.239; the Hanoi tap code date (1965) from POW memoirs as retold in news features; the newspaper-pricking story, which rests on secondary retellings; the 1950s date for reading Cryptic A, from news reports of the 2017 scroll.

The ancient world¶

Egypt: the scribe who changed the signs¶

Khnumhotep II, about 1900 BC¶

Enigmatic writing¶

Mesopotamia: a potter's recipe¶

The Hebrew Bible: Atbash¶

Swapping the alphabet end to end¶

The Dead Sea Scrolls¶

Greece: hidden messages¶

Herodotus's stories¶

Aeneas Tacticus: a chapter on secret messages¶

Aeneas's water clocks: a code book in a jar¶

Greece: the skytale¶

What Plutarch says¶

Was it a cipher at all?¶

Trying it¶

Greece: Polybius and his square¶

Torches on a wall¶

Rome: Caesar and Augustus¶

The shift of three¶

Greek letters on a spear¶

How strong is a shift?¶

Watchwords in the camp¶

China: tallies and split letters¶

India: secret writing as an art¶

Persia: the kings' secret script¶

Egypt after the pharaohs: Coptic ciphers¶

Can these be broken today?¶

Why weak ciphers worked¶

See also¶

References¶