Project Status: Active – The project has reached a stable, usable state and is being actively developed. Linux Build Status Windows Build status

High-level functions for supporting encryption and decryption of data from R. This allows secure storage and exchange of information, while trying to keep the encryption/decryption code from taking over your analyses. cyphr wraps the lower-level support from sodium and openssl. This package is designed to be easy to use, rather than the most secure thing (you’re using R, remember - for examples of what cyphr can’t protect against see jammr, rpwnd and evil.R.)

cyphr provides high level functions to:

The package aims to make encrypting and decrypting as easy as

cyphr::encrypt(save.csv(dat, "file.csv"), key)


dat <- cyphr::decrypt(read.csv("file.csv", stringsAsFactors = FALSE), key)

In addition, the package implements a workflow that allows a group to securely share data by encrypting it with a shared (“symmetric”) key that is in turn encrypted with each users ssh keys. The use case is a group of researchers who are collaborating on a dataset that cannot be made public, for example containing sensitive data. However, they have decided or need to store it in a setting that they are not 100% confident about the security of the data. So encrypt the data at each read/write.


To install cyphr from github:

remotes::install_github("ropensci/cyphr", upgrade = FALSE)


The scope of the package is to protect data that has been saved to disk. It is not designed to stop an attacker targeting the R process itself to determine the contents of sensitive data. The package does try to prevent you accidentally saving to disk the contents of sensitive information, including the keys that could decrypt such information.

Objects to handle keys:

Decide on a style of encryption and create a key object

cyphr does not include wrappers for key generation for sodium - sodium keys do not have a file format: So a secret symmetric key in sodium might be:

k <- sodium::keygen()
##  [1] 84 83 64 a2 3c 10 e6 9f 22 8b bc cd 48 81 0f 71 a0 4b 28 57 e3 40 c4
## [24] b7 ab 3e 00 f1 fd ff ee c8

With this key we can create the key_sodium object:

key <- cyphr::key_sodium(k)
## <cyphr_key: sodium>

If the key was saved to file that would work too:

If you load a password protected ssh key you will be prompted for your passphrase. cyphr will ensure that this is not echoed onto the console.

key <- cyphr::key_openssl()
## Please enter private key passphrase:

Encrypt / decrypt a file

If you have files that already exist and you want to encrypt or decrypt, the functions cyphr::encrypt_file and cyphr::decrypt_file will do that (these are workhorse functions that are used internally throughout the package)

saveRDS(iris, "myfile")
cyphr::encrypt_file("myfile", key, "myfile.encrypted")

The file is encrypted now:

## Error in readRDS("myfile.encrypted"): unknown input format

Decrypt the file and read it:

cyphr::decrypt_file("myfile.encrypted", key, "myfile.clear")
identical(readRDS("myfile.clear"), iris)
## [1] TRUE

Wrappers around R’s file functions

Encrypting files like the above risks leaving a cleartext (i.e., unencrypted) version around. If you want to wrap the output of something like write.csv or saveRDS you really have no choice but to write out the file first, encrypt it, and delete the clear version. Making sure that this happens even if a step fails is error prone and takes a surprising number of repetitive lines of code.

Alternatively, to encrypt the output of a file producing command, just wrap it in cyphr::encrypt

cyphr::encrypt(saveRDS(iris, "myfile.rds"), key)

Then to decrypt the a file to feed into a file consuming command, wrap it in cyphr::decrypt

dat <- cyphr::decrypt(readRDS("myfile.rds"), key)

The roundtrip preserves the data:

identical(dat, iris) # yay
## [1] TRUE

But without the key, it cannot be read:

## Error in readRDS("myfile.rds"): unknown input format

The above commands work through computing on the language, rewriting the readRDS and saveRDS commands. Commands for reading and writing tabular and plain text files (read.csv, readLines, etc) are also supported, and the way the rewriting is done is designed to be extensible.

The argument to the wrapped functions can be connection objects. In this case the actual command is written to a file and the contents of that file are encrypted and written to the connection. When reading/writing multiple objects from/to a single connection though, this is likely to go very badly.

Supporting additional functions

The functions supported so far are:

However, there are bound to be more functions that could be useful to add here (e.g., readxl::read_excel). Either pass the name of the file argument to cyphr::encrypt / cyphr::decrypt as

cyphr::decrypt(readxl::read_excel("myfile.xlsx"), key, file_arg = "path")

or register the function with the package using rewrite_register:

cyphr::rewrite_register("readxl", "read_excel", "path")

Then you can use

cyphr::decrypt(readxl::read_excel("myfile.xlsx"), key)

to decrypt the file (these are equivalent, but the former will likely be more convenient if you’re only dealing with a couple of files, the latter will be more convenient if you are dealing with many).

Collaborating with encrypted data

Even with high-level functions to ease encrypting and decrypting things given a key, there is some work to be done to distribute a set of keys across a group of people who are working together so that everyone can encrypt and decrypt the data but so that the keys themselves are not compromised.

The package contains support for a group of people are working on a sensitive data set. The data will be stored with a symmetric key. However, we never actually store the key directly, instead we’ll store a copy for each user that is encrypted with the user’s key. Any user with access to the data can authorise another user to access the data. This is described in more detail in the vignette (in R: vignette("data", package = "cyphr")).

Why are wrappers needed?

The low level functions in sodium and openssl work with raw data, for generality. Few users encounter raw vectors in their typical use of R, so these require serialisation. Most of the encryption involves a little extra random data (the “nonce” in sodium and similar additional pieces with openssl). These need storing with the data, and then separating from the data when decryption happens.


MIT © Rich FitzJohn.

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.