PNG Recon

Hide arbitrary data in plain site. Stuff files in PNGs and retrieve them later.

Optionally compress data.

Optionally encrypt data with a symmetric key derived from a salted password.

Install

Clone this repo.
Have virtualenv and python3
virtualenv -p python3 venv
source venv/bin/activate
pip install -I .
pngrecon -h

Usage

In general: use the encode command to stuff a file into a PNG. Use decode to retrieve it later.

By default, they both read from stdin and write to stdout.

Try pngrecon encode -h and pngrecon decode -h for a full list of options.

Example usage

Demonstrates data can be encoded and then decoded successfully.

(venv) user@host$ sha1sum README.md
1e2e72803fe66b8cdc2e7bfdf1eba6cf64584d2a  README.md
(venv) user@host$ <README.md pngrecon encode | pngrecon decode | sha1sum
1e2e72803fe66b8cdc2e7bfdf1eba6cf64584d2a  -

Common options

Pass -c gzip to pngrecon encode to compress encoded data with zlib.

# README.md is 1245 bytes but the data chunk storing it is only 615 bytes
(venv) user@host$ wc -c README.md
    1245 README.md
(venv) user@host$ <README.md pngrecon encode -c gzip | pngrecon info
/dev/stdin contains 5 chunks
Chunk IHDR with len 13
Chunk IDAT with len 10
ChunkType.Index with len 12
ChunkType.Data with len 615
Chunk IEND with len 0

With pngrecon encode, give -e to encrypt the data using a symmetric key. You may either give pngrecon a passphrase when prompted, or a --key-file from which to read a passphrase. Note: the entire contents of the --key-file will be used, including any trailing new line characters.

(venv) user@host$ echo -n "SuperSecurePassword" > pw.txt
(venv) user@host$ <file.txt pngrecon encode -e --key-file pw.txt | pngrecon decode --key-file pw.txt
[ ... contents of file.txt ... ]

Use -i and -o to change input/outout for encode and decode commands.

(venv) user@host$ pngrecon encode -i README.md -o hidden-readme.png
(venv) user@host$ file hidden-readme.png
hidden-readme.png: PNG image data, 1 x 1, 1-bit grayscale, non-interlaced

More examples

Encode all files in the current working directory with the help of tar.

(venv) user@host$ tar c . | pngrecon encode > hidden-backup.tar.png
(venv) user@host$ pngrecon info hidden-backup.tar.png
hidden-backup.tar.png contains 5 chunks
Chunk IHDR with len 13
Chunk IDAT with len 10
IndexChunk with len 12
DataChunk with len 10362880
Chunk IEND with len 0
(venv) user@host$ pngrecon decode -i hidden-backup.tar.png | tar t
[ ... list of files ... ]

Ideas

Add padding chunks.

When in the image, they should look exactly like a data chunk. So the data chunk type needs to be more complex.

Actually ... this seems quite hard. How big should the be? How many? Where in the file?

Warnings

Instead of relying on pngrecon to use standard cryptography libraries correctly, it might be better to use well-established and trusted encryption tools before piping data into pngrecon. We think we are using these libraries correctly, but if you need to rely on encryption, we think it is probably smarter to use something written by cryptographers.

We sought advice regarding compressing and encrypting data. The consensus seemed to be that compressing data and then encrypting the result is smarter than the opposite. That said, in certain situations, if your adversary can control part of the plaintext you are encrypting and can cause the encryption+compression to happen over and over again, they can learn the contents.

pastly/pngrecon