Earliest Hungarian Ciphers

In "Italian Cryptography (Stub)", I mentioned some ciphers related to Hungary in the archives of Modena and Milan described by Daniel Bourdeau :
(i) a cipher used in a letter in Latin (1482) of Matthias Corvinus, King of Hungary, to Ercole d'Este, Duke of Ferrara and his brother-in-law.
(ii) deciphered letters from Beatrice of Aragon, Matthias' queen, to her sister Elearnor, Duchess of Ferrara (1482, 1486).
(iii) a cipher used in letters in Italian (1482-1483) of Nicolò Sadoleto, Ferrarese envoy sent to Matthias Corvinus.
(v) a cipher used in letters (1489-1490) of Maffeo da Treviglio, Ludovico Sforza's agent at the court of Matthias Corvinus
(vi) another cipher used in a letter (1490) of Maffeo da Treviglio after the death of King Matthias.

In the above, (ii) and (iii) (from the State Archives of Modena) may be said to be Hungarian ciphers and I wondered how these are placed in Hungarian cryptography. I learned that these belong to the earliest period of Hungarian cryptography in:
Benedek Láng (2018), Real Life Cryptology: Ciphers and Secrets in Early Modern Hungary, Amsterdam University Press-Atlantis Press (Academia.edu) (in which many primary and secondary sources are given).

Hungarian cryptography began around the very reign of Matthias Corvinus (1458-1490) (p.136). Apart from (ii) the King's letter to the Duke of Ferrara and (iii) letters in the archives of Modena including those of Queen Beatrice (p.137, p.156, p.192), the following are mentioned:

A pigpen cipher (1470-1472) is recorded, with an example ciphertext of MATIAS [R]EX HUNGARIAE (p.136, 192). (This is by far older than other pigpen ciphers I know. See "Walsingham's Pigpen Cipher (1586-1587)".)

More "serious" ciphers were from Italy (p.137).

A cipher table prepared by Francesco Oliva for Matthias Corvinus (1488) is in the State Archives of Milan. It was homophonic, with graphical symbols. Its size is 140 and the vocabulary is in Italian (p.184). (His source is "Milano State Archives, Sf 650/3,5, photocopy: http://vestigia.hu/ ")

A cipher for ambassadors to Milan is also known. (p.137) (His source is "Péter E. Kovács (2003).)

There is also Beatrice's note in cipher written in a book in 1491 (after Matthias died and Vladislaus, Beatrice's second husband, was crowned) (p.137, 192).

From these, one may guess that Hungarians learned Italian ciphers from Ferrara and Milan.

The above ciphers are all for Latin and Italian. What about Hungarian? The first ciphertext in Hungarian with a definite date in Appendix is from 1531 ("Petrus Presbiter to Emerico de Putnok"). (p.193; See p.55, 57, 69, 171 for languages used in Hungary.)

 

The Gronsfeld Cipher before Count Gronsfeld - a Second Instance

Back in 2018, I wrote "The Gronsfeld Cipher before Count Gronsfeld". The cipher was for for Silvio Calandra (c.1540-1590), ambassador of the Duke of Mantua involved in diplomatic missions with Savoy and Milan in 1560s-1570s.

Now, I noticed a second instance for "Andreasi". Although the specific date is not known, it seems significant that this is again associated with Mantua.

See a new section "Gronsfeld Cipher" in "Italian Cryptography (Stub)".

Antipopes' Ciphers (1379-1412)

Gabriel de Lavinde's collection of ciphers (1379) preserved in the Vatican archives is well-known as early specimens of Western cryptography. I was inspired by Daniel Bourdeau's writing about a cipher of Avignon pope Benedict XIII to re-read the relevant section in Meister (1906) and learned Gabriel de Lavinde was affiliated with the antipope Clement VII in Avignon.

A natural question in discussing this topic is whether it formed the basis of later papal ciphers and whether it had anything to do with the pope's ciphers at Vatican at the time. Meister (1906) tells us that the papal ciphers in Avignon were more advanced than the primitive ciphers at Vatican but ceased to be used with the end of the Avignon papacy.

See my new article, "Papal Ciphers during the Western Schisma".

 

Anthropic's Claude Finds Origin of Parameters for Elliptic Curve Encryption

For a change, it's about modern cryptography.

AI not only breaks or otherwise deciphers historical ciphers. Recently, Anthropic's Claude identified the origin of parameters used in elliptic curve cryptography. This does not mean the widely used encryption scheme is compromised. To the contrary, it alleviates the concern that the system parameters might have been intentionally chosen to provide a backdoor for the NSA.

First, I thought of reporting this in a few lines, but while reading the announcement, I felt like adding some basics for people (like me) not versed in computing/programming.

See my new article, "Anthropic's Claude Identifies Origin of Seed for Elliptic Curve Encryption".

 

Italian Cryptography

I uploaded a new article, "Italian Cryptography (Stub)".

This is like a note of references that I should read but in collecting these from my memo, I realized how homophonic substitution was already common early in the fifteenth century. And it is impressive that a 1424 Milanese cipher already had representation of syllables.

A Numerical Cipher of Japanese Army (1902-1904) Reconstructed from Japanese Archives

I uploaded a new article, "A Numerical Cipher of the Imperial Japanese Army (1902-1904)", which describes one numerical cipher of the Imperial Japanese Army in the Meiji Era reconstructed by Harumi-Matsuba.

The relevant materials were found in JACAR (Japan Center for Asian Historical Records). Although I worked on this database more than ten years ago, I didn't notice this. Come to think of it, I think my scope then was limited to what were found with the search term 暗号 (code/cipher), which already returned quite a few hits.

As noted yesterday, recent studies found a substantial collection of cipher materials related to Sun Yat-sen that escaped my notice. Hopefully, there will be a lot more to be found about Japanese (and Chinese) ciphers from this database.

Three Mingmi Codes Related to Sun Yat-sen

The other day, I reported Daniel Bourdeau solved and deciphered many telegrams related to Sun Yat-sen. Now Brice Dudout are discovering more keys and deciphering many more telegrams. Among them, there are three indicated with "X密", which scramble codebook digits rather than converting them into Roman letters.

For two of them, Dudout identified scrambling formulae combining subtraction/addition with table mapping. I now suggest they can be reformulated as simple renumbering of pages/columns/rows in the mingmi style codebook as in the case of the third. I described this in a new section, "Mingmi (明密) Format Code: Gongmi (共密), Rongmi (容密), Chenmi (辰密)", here.

 

 

Emperor Maximilian's Cipher from 1496?

I uploaded a new article, "An Imperial Cipher from 1496?".

This is about a cipher recently identified independently by Daniel Bourdeau and George Lasry. Lasry dated it to 1496, to which Bourdeau concurs. If this is indeed from 1496, this may be related to the beginning of use of symbol ciphers by the Habsburgs.

My article concludes with two desiderata. First, crucial evidence of the dating seems to be G. A. Pecci (1755), but I could not see the correspondence. Secondly, if the cipher is shared with Sienese, the Habsburgs may have learned the cipher from them. I wonder whether the letter is really addressed to Sienese envoys.

If someone can find out anything, I'd like to know.

Verification of AI and Verification by AI

I uploaded a new article, "Verification of AI Decipherment and Verification by AI".

It first deals with how one can verify a decipherment. This is a matter of course for crypto-people but I recently received an inquiry about this. The case of symbols occurring only once that I mentioned yesterday is worth some consideration in this regard.

The latter half gives techniques AI use to verify their own results. This is something I read in many AI solutions that have emerged these days.

Empress Eugénie's Cipher Telegrams to Napoleon III Broken by Two

Again, two people independently achieved codebreaking with AI almost at the same time. They decoded telegrams between Napoleon III and Empress Eugénie in June 1859.

Geoge Lasry notified me of his solution (one in a batch of no less than 56 solutions) on 10 October 2026 and I published his results here on 12 October 2026. Jimmy Apples had published his results on X on 5 October 2026 and provided me with details on 13 October 2026. I now mentioned him in my report of George's results.

Two of the decoded telegrams are from Eugénie to Napoleon and the others are from the Emperor to the Empress.

Eugénie's telegram on 23 June 1859 reads:

Je suis désolée de vous avoir fait de la peine. Ce n'était pas mon intention. Ne vousinquiétez pas du petit, il va à merveille. J'écris aujourd'hui par la poste.
(I am so sorry to have caused you pain. It was not my intention. Do not worry about the little one, he is doing wonderfully. I am writing today by post. (Apples' translation))

The short telegram of the Empress is about personal matters, though the Empress acted as regent during the Emperor's absences in 1859, 1865, and 1870 (Wikipedia). In 1859, Napoleon III was with the army in the Italian campaign, which culminated in the Battle of Solferino on 24 June.

In terms of cryptanalysis, this is interesting in two points.

(1) No plaintext is found and this is a ciphertext-only achievement.

(2) One telegram has a sequence in which many figures occur only once:

677(je) 1074(re) 310 925 357(des) 381 708(le) 81 1099 123 495 656 257 914(par) 439(ennemi)

Apples succeeded in filling these:

677(je) 1074(re) 310([cueille]) 925([partout]) 357(des) 381([do]) 708(le) 81([ances]) 1099([relatives]) 123([aux]) 495([exact]) 656([ions]) 257([commises]) 914(par) 439(ennemi)

The readings identified for these code groups fit in appropriate places in the alphabetically ordered code list. So this is not just a fabrication from nowhere. But it is not straightforward to find out whether this is the only candidate.

 

Dutch Ciphers with Table Switching

It has been known from Karl de Leeuw's work that Dutch diplomatic ciphers provided multiple substitution tables.

Now, in deciphering a ciphertext in Thurloe State Papers, Paolo Rosson found it switches among five tables in one letter. Though the plaintext was found in a Dutch publication, the scheme he found out is interesting.

The ciphertext is something like "113. 10. 26. 13. 13. 19. 29. 29. 26. 11. 222. 31. 17. 21. 22. 23. 23. 22.13. 30. 20. 18. 22. 23. 15. 13. 24. 22. 13. 24. 22. 17. 10. 15. 25. 23. 13. 329. 28. 15. 22. 20. 25. 20. 15. 25. 27. 28. 20. 27. 27. 20. 22. 17. 20. 15. 527. 16. 20. 21. 12. 19. 21. 29. 26. 20. 16. 27. 16. 19...." Although the specimen is a bit short, I thought solving this would not be difficult.

As it turned out, this short letter switches tables multiple times. No wonder it couldn't be solved with ordinary attacks. (I said the plaintext might not be in Dutch but in French. Some may have been misled by this.)

The first digit of three-digit group indicates one of tables 1-5.

I uploaded a new article about this, "Table Switching in Dutch Diplomatic Ciphers".

Trithemius' Cipher Deciphered and Re-enciphered Polyalphabetically in the 17th Century (Heidel Cipher)

It is well-known that the cipher in the third book of Trithemius' Steganographia was solved in the 1990s. But it was only recently that I learned the cipher was broken in the seventeenth century by one Wolfgang Ernst Heidel, who declared it in his 1676 book but re-enciphered his solution in his own polyalphabetic cipher!

Thomas Ernst, who had solved Trithemius' cipher, deciphered Heidel's cipher and thus proved that Heidel indeed had the solution of Trithemius' cipher. Since Heidel's cipher had apparent word boundaries, a breakthrough was made when he identified the matching ciphertext for "gaza frequens libicos carthago duxit triumphos" known from the solution of Trithemius' cipher.

Thomas Ernst had succeeded in deciphering Trithemius cipher in 1993 and published it in 1996. In 1998, Jim Reeds independently solved it and then found it was already solved before him. Cryptologia (1998, issue 4) carried Jim Reeds' paper together with Thomas Ernst's brief English summary of his original paper in German.

Since it may take time before I have time to study further details, references are given below:

Thomas Ernst, "Schwarzweiße Magie: Der Schlüssel zum dritten Buch der Steganographia des Trithemius", Daphnis: Zeitschrift für Mittlere Deutsche Literatur, 25/1 (1996), pp. 1-205

Thomas Ernst, "The Numerical-Astrological Ciphers in the Third Book of Trithemius’s Steganographia", Cryptologia, XXII/4 (1998), pp. 318-341

Jim Reeds, "Solved: the Ciphers in Book III of Trithemius's Steganographia", Cryptologia, XXII/4 (1998), pp.291-317

Thomas Ernst's information about Heidel is on a wordpress-site: Restoring the Steganographia of Trithemius

Thomas Ernst analysed a forged manuscript of the "Steganographia" presumedly from 1521: "Anatomie einer Fälschung: ‘Johannis Trithemij [...] Steganographiæ Lib· 3. cum Clave, tàm generalj, quàm specialj [...] M.D.XXI’", Daphnis: Zeitschrift für Mittlere Deutsche Literatur und Kultur der Frühen Neuzeit (1400–1750), 30/3-4 (2001), pp. 513-595

"Hieronimo Ranzo" Cipher Solved by AI

A cipher with superscripts with word elements are now solved by Paolo Rosson. A few hours after his email, I received a report of solution from George Lasry.

This is what I listed as "Venetian? Cipher with Superscript Digits (1528)". As it turned out, the writer was a Genoese with the pseudoonym "Hieronimo Ranzo", apparently different from "Hieronimo Ranzo" known in history. Only yesterday, I gave this one that remains intractable by AI!

I uploaded a new article about this, "A Cipher with Superscripts for Word Elements ("Hieronimo Ranzo") Solved by AI".

 

Anatomy of AI Codebreaking

I uploaded a new article, "Codebreaking with AI".

In the past month, I received many solutions for pieces in my list of unsolved ciphers. I suppose most of them were achieved with the assistane of AI.

Although I'm not versed in technical details, I tried to sort out what AI did in these achievements.

Possibly Manual Solution of a Very Short Numerical Ciphertext: Ormonde-Maltravers Cipher (1634-1635)

While Daniel Bourdeau is known to have solved many ciphers by using AI, his simulated-annealing homophonic solver didn't work for Ormonde-Maltravers Cipher (1634-1635). The ciphertext looks too short for ordinary algorithmic analysis. His solution seems more like human ingenuity. It can be seen as an interesting example of good old codebreaking. (Github records are difficult for me to decipher, and I don't know how far he was helped by AI in this.)

I uploaded a new article to describe his solution: "Ormonde-Maltravers Cipher (1634-1635)".

Did François Viète Break a Polyphonic Cipher?

I reported Paolo Rosson's full decipherment of two letters in a polyphonic cipher in a new article, "An Intercepted Report of Henry IV's Upcoming Attendance at Mass in a Polyphonic Cipher". When I identified the key from deciphered materials back in 2017, I reported these as undeciphered. The fresh decipherment reveals reporting of the prospect of Henry IV's attending Mass.

Rosson noted that these were among the intercepted letters but were recorded as left undeciphered. Probably, because the letters to the Pope and the Bishop of Lizieux from the same date were deciphered, deciphering these was not considered important. Since Rosson provides the full text, I hope someone will compare it with the known letters to see if there is some new content. (Maybe AI can do such a specific task?)

From the viewpoint of cryptography, it is wondered who deciphered the letters to the Pope and the bishop. It is exciting to speculate whether it was François Viète.

 

Enigma Messages Solved by AI

A wave of codebreaking by/with AI now hit unsolved Enigma messages. Three people independently reported solutions of Enigma messages on a list of Enigma messages in Frode Weierud's CryptoCellar. (Some messages used for his coauthored paper (Olaf Ostwald and Frode Weierud (2017), "Modern breaking of Enigma ciphertexts" (Cryptologia)) are here.)

(1) A message dated 10 July 1941 with a discriminant (Kenngruppe) MVUEH was broken by Carter Leffen (X), who used OpenAI GPT-6 Astra. He communicated his solution to Weierud on 15 September 2026 (see Weierud's report).

As with the case of Vals AI's solution of Cyphral Distich reported yesterday, Leffen only told AI to see if it could break any of the unbroken Enigma messages on the CryptoCellar web page. All by itself, AI picked up Nr.172 (MVUEH) and assumed that its content might be related to the known plaintext of Nr.173 (SIPVX) from the same date. After many failed attempts, AI took up the repeated place name ROSENOW ROSENOW as a crib. AI then autonomously developed necessary Python and C++ software for an Enigma simulator and an Enigma Bombe and finally it successfully identified the key. Its plaintext turned out to be almost identical with Nr.173 (SIPVX). Actually, the SIPVX key had been tried before but it didn't work. As it turned out, the MVUEH key is completely different from the SIPVX key.

One more amazing thing about this solution gleaned from the log is that AI discovered relevant documents in the archives by itself, though it is not yet clear how far AI used the archival materials.

(2) Solution of Nr.285 (FMNGI), dated 31 July 1941, by Jack Willis using Anthropic's Claude Opus 5 was reported to Weierud on 21 September 2021. Unlike the case of MVUEH, this solution was "more directed and relied on strong human guidance" (see Weierud's report).

AI made the transcription of the message and used an already broken message ALQFI between the same two stations as a control to check the transcription.

Several messages (including ALQFI) from the same sender suggested a crib XHARTJENSTEINX (the sender was SS-Obersturmbannführer, Friedrich Hartjenstein). Variants of this name had already been given in Geoff Sullivan and Frode Weierud (2005), "Breaking German Army Ciphers", Cryptologia (pdf).

As it happened, Weierud had found a plaintext in July 2026 and put a notice of it on his website but not the plaintext itself. So the solutioin did not use the known plaintext.

(3) Solution of messages Nr.187 (AWTZK) and Nr.189 (ZNLZT) from 11 July 1941 by John Woodhouse (pdf) using Anthropic's Claude was reported to Weierud on 26 September 2026 (Weierud's report).

Again, a crib was the key to the success. Twelve cribs, harvested from messages with close dates, were "tested with a Bombe-style search that tolerates one garbled letter." The fifth crib XOSTROWXOSTROWX (again, a repeated place name) succeeded for ZNLST, and the key obtained therefrom could read AWTZK.

The search was helped by an observation from 1941 key sheets that "no wheel sits in the same slot on cnsecutive days". This reduced 60(=5*4*3) wheel orders to 14 for 11 July because 10 July employed V-II-I and 12 July emplyoed II-V-IV.

(By the way, I learned of this topic by newsfeeds in Japanese: GIGAZINE, netlife.)

AI's Solution of Urquhart's Cyphral Distich Verified

A short numerical ciphertext called Cyphral Distich in an edition of Logopandecteision by Sir Thomas Urquhart was solved by AI (Claude Fable 5.1), as announced by Geby Jaff (Vals AI) on 31 August 2026.

It is remarkable that Fable 5.1 was simply told to solve an unsolved cryptogram (one that is deterministic and not ultra-hard) and autonomously picked up this and solved it without any human intervention. What looks like an ordinary numerical cipher turned out to be a book cipher.

Right after the announcement, sceptics pointed out that the ciphertext is not present in the 1653 edition available online and the decryption could not be reproduced.

Actually, the ciphertext is found at least in an 1834 edition and using its text (different from that of the 1653 edition) reproduces the decryption. This was pointed out at least by Jozef Krajčovič（facebook, 5 September 2026), kjfriedo (comments on Bruce Schneier's blog, 9 September 2026), AI時短ラボ (17 September 2026). But because sceptic views continued to be professed, I prepared my own verification here, and uploaded a new article in Japanese, "AIが人間の介在なしに17世紀の暗号Cyphral Distichを解読したのは本物だった".

Urquhart also left a second ciphertext called Cyphral Octastich and solved it in a similar way. Daniel Bourdeau, who also used Fable 5.1 to solve many ciphers including those on my list, verified the solution. He independently applied the rule found by Vals AI to reproduce the announced reading, though some numbers were "one word off".

kjfriedo (comments on Bruce Schneier's blog, 10 September 2026) had also partially verified the solution.

Walsingham's Pigpen Cipher (1586-1587)

I uploaded a new article about a pigpen cipher used in letters from Thomas Wilkes to Sir Francis Walsingham. See "Walsingham's Pigpen Cipher (1586-1587)"

I wrote the most part of this some weeks ago but I didn't have time to complete it until now.

 

A Numerical Cipher in an Italian Letter to Francis I (1530)

During the sixteenth century, numerical ciphers were not common in France. One letter from London to Francis I (1530) is the earliest French numerical cipher I know, though, as expected, the letter is in Italian.

For a long time, I have been wondering what kind of cipher this is. Now Daniel Bourdeau deciphered it. I uploaded a new article about this, "A Numerical Cipher in a Letter to Francis I (1530)".