"FontCode" which can hide secret messages in the document at a level that can not be confirmed with the naked eye

Known as invisible ink unless special treatment is givenInvisible ink"Mankind has developed many means for exchanging confidential information that should not be seen by others. Research teams such as Associate Professor Changsi Dun of Columbia University have developed a technology to embed secret messages that can not be confirmed with the naked eye in the document "FontCodeSucceeded in the development of.

[1707.09418] FontCode: Embedding Information in Text Documents using Glyph Perturbation
https://arxiv.org/abs/1707.09418

Researchers hide information in plain text
https://techxplore.com/news/2018-05-plain-text.html

Regarding fields where FontCode is likely to be useful, Mr. Zhun says, "As well as the uses related to spy activities, companies that want to prevent document alteration and copyright protection, and retailers and artists do not change the appearance and layout of documents IntoQR codeThere are cases where you might want to embed metadata or metadata. " In order to realize these needs,SteganographyWe have succeeded in realizing "FontCode" which embeds information such as text data and metadata in the document.

FontCode changes the parts such as the thickness of the line used in the font, the skein, and the degree of bending, as follows, and assigns numbers such as "1" "2" "3" to each part. Since this change can prepare multiple patterns for one character, it is possible to have multiple numbers for one letter "a", for example.



It is difficult to grasp the difference of fonts with the naked eye ... ...


It is possible to recognize if it is a computer, and you can easily retrieve messages hidden in the document.


Also, when embedding a secret message in a document, FontCode converts the message entered by the user into a number. After that, it is possible to embed the information by extracting the place (pink color) where the font is changed from the document and converting it into the font matching the number.


FontCode can also encrypt the message so that it can not be read by anyone other than a specific person. Numbers and fonts are linked one to one as follows, but by replacing this association it becomes possible to embed information in an encrypted state.


For this reason, in order to read a secret message, the same "font and number" correspondence table (Common keyThe possibility of being read by a third party is low.


A message embedded by FontCode can be read if it can recognize even the characters of the changed font. For this reason, documents embedded with FontCode have a feature that they do not depend on the storage form of paper medium, image file, PDF file, etc. In addition, Mr. Zhun says that "general versatility is high" because it corresponds to general things such as "Times Roman" "Helvetica" "Calibri" as fonts available for encryption.

Mr. Zun 's research team is supposed to shoot from posters and documents embedded with texts such as URLs in FontCode using smartphone cameras like QR code for use. However, depending on the angle of shooting and lighting etc., it is impossible to read detailed font changes, and it is pointed out that there are cases where the embedded information can not be read accurately.

So, to rescue unreadable characters,Chinese Remainder TheoremWe devise a method to guess using. This means that the original value can be derived from a plurality of remainder values, such as "a remainder obtained by dividing a certain value by 3, a remainder obtained by dividing by 5, a remainder obtained by dividing by 7" In FontCode, we implemented a mechanism to derive missing characters using this theorem, making it possible to extract the original message even in a state where up to 25% of characters can not be recognized.

Mr. Zhun says, "FontCode is compatible with alphanumeric characters only, but we plan to expand it to support other languages ​​including Chinese."

The structure of "FontCode" is also explained in movies published on YouTube.

FontCode: Embedding Information in Text Documents using Glyph Perturbation - YouTube