The “X密” books: one code per correspondent

Satoshi Tomokiyo’s idea

We had described our three numeric keys with formulae: a subtraction for 共密, substituted digits plus an addition for 容密, pages counted backwards without zeros for 辰密. On 23 October 2026 Satoshi Tomokiyo (Cryptiana website) proposed reading them as the 1916 clerk used them: with a copy of the “明密” telegraph code whose page, column and row numbers had been rewritten by hand^[22]. He showed it on a June 1916 copy in his possession.

We checked it: over all 10,000 possible groups, our three keys reduce exactly to three renumberings, one for pages, one for columns, one for rows. No carry is left: what we called a “carry” was only the xxx0 character, placed at the foot of the previous column. Only one page diverged, and the one group concerned, in N25, is in fact two swapped digits. Corrected, it gives 一切. Every key page on the site now points to his idea.

One book per correspondent

Huang Xing’s drafts published for the summer of 1916 name the code used at the head of each telegram^[29]. They show 嚴密 for Peng Yunyi (彭允彝) in Beijing, 辰密 for Zeng Jiwu (曾繼梧) in Changsha, and 共密 for Chen Jiahui (陳嘉會) and Li Shucheng (李書城). Then come 仁密 for Gu Zhongxiu (谷鍾秀), 信密 and 安密 for Tan Yankai (譚延闓), and 箴密 for Sun Hongyi (孫洪伊). Satoshi Tomokiyo also notes that many “X密” books of 1913–1923 are preserved in Taiwan^[22].

His idea explains the system. Each correspondent received the same printed book, but with his own numbers, rewritten by hand. Handing out a key cost only one copy and a few hours of copying. In return, all the secrecy lay in that renumbering: once the principle is understood, each book breaks like the others.

乾密: the telegram of the day after

So we applied the model to the “X密” telegrams that were still unreadable, with controls fixed before computing. One of them opens: N09, prefixed 乾密, sent from Huang Xing’s house in Tokyo (青山南町三ノ五三). The Japanese copy is undated, but it is filed between the telegrams of 6 and 9 June^[27].

袁死。此間態度未變，候交涉有望，再電告。興。
“Yuan is dead. Here the attitude has not changed; I shall wait until the negotiations look promising and cable again. Xing.”

This is the first reaction of Huang Xing to Yuan Shikai’s death (6 June) that we have. It is sober: the news changes nothing in Tokyo, and he will contact his correspondents again only with something concrete. The “negotiations” are probably those told in the 容密 telegrams: arms and loans in Japan for the provinces in revolt. What follows bears this out. On 9 June, in N13, he writes that “Japan is still waiting and seeing”; on the 10th, in N17, that the Japanese authorities will not lend to Beijing before North and South are unified.

The address, in clear, asks for the message to be passed to “Ma and Geng, my two brothers”. They are not identified. For Geng, one lead: in a letter of 1 June 1916 published with the drafts, Sun Hongyi says he conferred with Zhang Ji, Bo Wenwei, 耿鶴生 and He Chengjun (何成濬), “men of the party who have forces”^[29]. The Japanese addressee, ジヨンシヨウ, lives, like the Hunanese group, on Avenue Joffre (霞飛路) in the French concession.

The Japanese Ministry had N01’s key

In the same file, a sheet of the Japanese Foreign Ministry from June 1916 carries two tables of digits, place by place, one to encipher and one to decipher^[27]. We did not know what they were for. They read N01, an appeal from Shanghai to Huang Xing received in May: “袁逆將倒，時局益艱……盼承指導，俾有遵循” (“Yuan is about to fall, the situation grows tense… we await your guidance”), signed with some fifteen characters (建 for Niu Yongjian, 闓 for Tan Yankai, 秀 for Gu Zhongxiu, 驪 for Yin Ruli…). Better still: the key our search had found for N01, with Satoshi Tomokiyo’s model and without these tables, is identical to the Ministry’s. Japan was therefore reading this channel in 1916; we do not know whether it broke or obtained the key. See the method.

What still resists

• N03, another 乾密 telegram sent from the same house in early June, stays unreadable with N09’s key: another day, another key, or faulty digits.
• N06 (和密, eleven groups) is too short; N07 does not open with this model.

Further progress needs other telegrams in the same books. The collections in Taiwan mentioned by Satoshi Tomokiyo, or the books themselves, would be the best way.

Caution

N09’s key rests on a single telegram of 17 groups. It passes our checks: a score well above the controls, the expected signature and consistency with the date. But three characters remain probable readings: 未, 候 and 告. The digits were re-read twice on the image by Brice Dudout, the second time blind. See the method and the 乾密 page.