paperbackup.py

Create a pdf with barcodes to backup text files on paper. Designed to backup ASCII-armored GnuPG and SSH key files and ciphertext.

How to use

Backup

gpg2 --armor --export-options export-minimal --export-secret-key "User Name" >key.asc
paperbackup.py key.asc
paperrestore.sh key.asc.pdf | diff key.asc -
lpr key.asc.pdf

This will print out the public and private key of "User Name". The private key is still encrypted with it's passphrase, so make sure you don't lose or forget it.

See some example output here: https://github.com/intra2net/paperbackup/raw/master/example_output.pdf

Restore

1. Scan the papers
2. Create one file containing all the pages. zbar supports e.g. PDF, TIFF, PNG, JPG,...
3. paperrestore.sh scanned.pdf >key.asc
4. gpg2 --import key.asc

If one or more barcodes could not be decoded, try scanning them again. If that does not work, type in the missing letters from the plaintext output at the end of the pdf.

Dependencies

• python 3 https://www.python.org/
• python3-pillow https://python-pillow.org/
• PyX http://pyx.sourceforge.net/
• LaTeX (required by PyX) https://www.latex-project.org/
• python3-qrencode https://github.com/Arachnid/pyqrencode
• enscript https://www.gnu.org/software/enscript/
• ghostscript https://www.ghostscript.com/
• ZBar http://zbar.sourceforge.net/

Why backup on paper?

Some data, like GnuPG or SSH keys, can be really really important for you, like that your whole business relies on them. If that is the case, you should have multiple backups at multiple places of it.

I also think it is a good idea to use different media types for it. Hard disks, flash based media and CD-R are not only susceptible to heat, water and strong EM waves, but also age.

Paper, if properly stored, has proven to be able to be legible after centuries. It is also quite resistant to fire if stored as a thick stack like a book.

So I think it is a good idea to throw a backup on paper into the mix of locations and media types of your important backups.

Storing the paper backup in a machine readable format like barcodes makes it practical to restore even large amounts in short order. If the paper is too damaged for the barcodes to be readable, you still have the printed plaintext that paperbackup produces.

How to properly store the paper

The ISO has some standards for preservation and long term storage of paper:

ISO/TC 46/SC 10 - Requirements for document storage and conditions for preservation http://www.iso.org/iso/home/store/catalogue_tc/catalogue_tc_browse.htm?commid=48842

Here's an example of what ISO 16245 describes: http://www.iso.org/iso/livelinkgetfile-isocs?nodeId=15011261

Choice and error resilency of barcodes

Only 2D barcodes have the density to make key backup practical. QR Code and DataMatrix are the most common 2D barcodes.

Using a common barcode symbology makes sure that there are several independent implementations of decoders available. This increases the probability that they handle defects and error correction differently and are able to tolerate different kinds of defects. So if the barcode gets damaged, you have several programs you can try.

Several papers comparing QR and DataMatrix come to the conclusion that DataMatrix allows a higher density and offers better means for error correction. I tested this and came to the conclusion that the QR code decoding programs available to me had better error resilency than the ones for DataMatrix.

The toughest test I found, other than cutting complete parts from a code, was printing the code, scanning it, printing the scanned image on a pure black and white printer and then repeating this several times. While the barcode still looks good to the human eye, this process slightly deforms the barcode in an irregular pattern.

libdmtx was still able to decode a DataMatrix barcode with 3 repetitions of the above procedure. A expensive commercial library was still able to decode after 5 repetitions.

ZBar and the commercial library could still decode a QR code after 7 repetitions.

A laser printed QR code, completely soaked in dirty water for a few hours, rinsed with clean water, dried and then scanned, could be decoded by ZBar on the first try.

This is why I chose QR code for this program.

Encoding and data format

In my tests I found that larger QR codes are more at risk to becoming undecodable due to wrinkles and deformations of the paper. So paperbackup splits the barcodes at 140 bytes of data.

QR codes offer a feature to concatenate the data of several barcodes. As this is not supported by all programs, I chose not to use it.

Each barcode is labeled with a start marker ^<sequence number><space>. After that the raw and otherwise unencoded data follows.

Plaintext output

paperbackup prints the plaintext in addition to the QR codes. If decoding one or more barcodes should fail, you can use it as fallback.

To ease entering large amounts of "gibberish" like base64 data, each line is printed with a checksum. The checksum is the first 6 hexadecimal characters of MD5 sum of the line content. The MD5 is on the "pure" line content without the line break (e.g. \n or \r\n)

To verify a line checksum use echo -n "line content" | md5sum | cut -c -6

If a line is too long to be printed on paper, it is split. This is denoted by a "^" character at the begin of the next line on paper. The "^" is not included in the checksum.

Changing the paper format

The program writes PDFs in A4 by default. You can uncomment the respective lines in the constants section of the source to change to US Letter.

Similar projects

paperbackup with reportlab backend https://github.com/tuxlifan/paperbackup

Should behave the same as this paperbackup but with using reportlab instead of PyX/LaTeX for PDF generation. Any discrepancies should be filed at https://github.com/tuxlifan/paperbackup/issues

PaperBack http://ollydbg.de/Paperbak/ and https://github.com/Rupan/paperbak/

Although it is GPL 3, this original version PaperBack (program 8-character name PaperBak) is for Windows only (but in 2018 a crossplatform, backwards-compatible, command line version paperbak-cli has been published, see next entry). It uses it's own proprietary barcode type. That allows it to produce much more dense code, but in case of a problem with decoding you are on your own.

paperback-cli https://git.teknik.io/scuti/paperback-cli

Paperback-cli is the crossplatform, backwards-compatible, command line version of Oleh Yuschuk's PaperBack. https://github.com/Wikinaut/paperback-cli is a copy on github. See discussion on Rupan/paperbak#1 for further programs.

ColorSafe https://github.com/colorsafe/colorsafe

A data matrix for printing on paper. Inspired by PaperBak, ColorSafe is written with modern methods and technologies and is cross-platform. It aims to allow a few Megabytes of data (or more) to be stored on paper for a worst case scenario backup, for extremely long-term archiving, or just for fun. With best practices, ColorSafe encoded data can safely withstand the viccissitudes of technology changes over long periods of time.

Twibright Optar http://ronja.twibright.com/optar/

Uses the not-so-common Golay code to backup 200KB per page. So it offers a much higher density than paperbackup.py, but is probably more affected by defects on the paper. GPL 2 and designed for Linux.

Paperkey http://www.jabberwocky.com/software/paperkey/

It is designed to reduce the data needed to backup a private GnuPG key. It does not help you to print and scan the data. So it could be used in addition to paperbackup.py.

asc2qr.sh https://github.com/4bitfocus/asc-key-to-qr-code

Very similar to paperbackup.py. But it only outputs .png images without ordering information. So you have to arrange printing and ordering yourself.

License

MIT X11 License