Skip to contents
 /\ sec   \
/  \ ret   \
\  /  base /

Fast and memory-efficient streaming hash functions and base64 encoding and decoding.

Performs direct hashing of strings and raw vectors. Stream hashes files potentially larger than memory, as well as in-memory objects through R’s serialization mechanism.

Implementations include the SHA-256, SHA-3 and ‘Keccak’ cryptographic hash functions, SHAKE256 extendable-output function (XOF), and ‘SipHash’ pseudo-random function.

The SHA-3 Secure Hash Standard was published by the National Institute of Standards and Technology (NIST) in 2015 at doi:10.6028/NIST.FIPS.202. SHA-3 is based on the Keccak algorithm, designed by G. Bertoni, J. Daemen, M. Peeters and G. Van Assche.

The SHA-256 Secure Hash Standard was published by NIST in 2002 at

The SipHash family of pseudo-random functions by Jean-Philippe Aumasson and Daniel J. Bernstein was published in 2012 at[1]



For the SHA-3 cryptographic hash algorithm, specify ‘bits’ as 224, 256, 384 or 512:

sha3("secret base")
#> [1] "a721d57570e7ce366adee2fccbe9770723c6e3622549c31c7cab9dbb4a795520"
sha3("secret base", convert = FALSE)
#>  [1] a7 21 d5 75 70 e7 ce 36 6a de e2 fc cb e9 77 07 23 c6 e3 62 25 49 c3 1c 7c
#> [26] ab 9d bb 4a 79 55 20
sha3("秘密の基地の中", bits = 512L)
#> [1] "e30cdc73f6575c40d55b5edc8eb4f97940f5ca491640b41612e02a05f3e59dd9c6c33f601d8d7a8e2ca0504b8c22f7bc69fa8f10d7c01aab392781ff4ae1e610"

Hash strings and raw vectors

Character strings and raw vectors are hashed directly (as per the above).

Stream hash R objects

All other objects are stream hashed using R serialization

  • memory-efficient as performed without allocation of the serialized object
  • portable as always uses R serialization version 3 big-endian representation, skipping headers (which contain R version and native encoding information)
sha3(data.frame(a = 1, b = 2), bits = 224L)
#> [1] "03778aad53bff7dd68caab94374bba6f07cea235fb97b3c52cf612e9"
#> [1] "b3e37e4c5def1bfb2841b79ef8503b83d1fed46836b5b913d7c16de92966dcee"

Stream hash files

Files are read and hashed incrementally, accepting files larger than memory:

file <- tempfile(); cat("secret base", file = file)
sha3(file = file)
#> [1] "a721d57570e7ce366adee2fccbe9770723c6e3622549c31c7cab9dbb4a795520"

Hash to integer / SHAKE256 XOF

May be used as deterministic random seeds for R’s pseudo random number generators (RNGs).
Specify ‘convert’ as NA (and ‘bits’ as 32 for a single integer value):

shake256("秘密の基地の中", bits = 32L, convert = NA)
#> [1] 2000208511

For use in parallel computing, this is a valid method for reducing to a negligible probability that RNGs in each process may overlap. This may be especially suitable when first-best alternatives such as using recursive streams are too expensive or unable to preserve reproducibility. [2]


keccak("secret base", bits = 384L)
#> [1] "c82bae24175676028e44aa08b9e2424311847adb0b071c68c7ea47edf049b0e935ddd2fc7c499333bccc08c7eb7b1203"


sha256("secret base")
#> [1] "1951c1ca3d50e95e6ede2b1c26fefd0f0e8eba1e51a837f8ccefb583a2b686fe"

For a SHA-256 HMAC, pass a character string or raw vector to ‘key’:

sha256("secret base", key = "秘密の基地の中")
#> [1] "ec58099ab21325e792bef8f1aafc0a70e1a7227463cfc410931112705d753392"


SipHash-1-3 is optimized for performance.
Pass a character string or raw vector to ‘key’ - up to 16 bytes (128 bits) of the key data is used:

siphash13("secret base", key = charToRaw("秘密の基地の中"))
#> [1] "a1f0a751892cc7dd"

Base64 Encoding / Decoding


base64enc("secret base")
#> [1] "c2VjcmV0IGJhc2U="
base64dec(base64enc("secret base"))
#> [1] "secret base"

Raw vectors:

base64enc(as.raw(c(1L, 2L, 4L)), convert = FALSE)
#> [1] 41 51 49 45
base64dec(base64enc(as.raw(c(1L, 2L, 4L))), convert = FALSE)
#> [1] 01 02 04

Serialized objects:

base64dec(base64enc(data.frame()), convert = NA)
#> data frame with 0 columns and 0 rows


Install the latest release from CRAN or R-multiverse:


The current development version is available from R-universe:

install.packages("secretbase", repos = "")

Implementation Notes

The SHA-256, SHA-3, Keccak, and base64 implementations are based on those by the ‘Mbed TLS’ Trusted Firmware Project at

The SipHash implementation is based on that of Daniele Nicolodi, David Rheinsberg and Tom Gundersen at, which is in turn based on the reference implementation by Jean-Philippe Aumasson and Daniel J. Bernstein released to the public domain at


[1] Jean-Philippe Aumasson and Daniel J. Bernstein (2012), “SipHash: a fast short-input PRF”, Paper 2012/351, Cryptology ePrint Archive,

[2] Pierre L’Ecuyer, David Munger, Boris Oreshkin and Richard Simard (2017), “Random numbers for parallel computers: Requirements and methods, with emphasis on GPUs”, Mathematics and Computers in Simulation, Vol. 135, May 2017, pp. 3-17 doi:10.1016/j.matcom.2016.05.00.

◈ secretbase R package:

Mbed TLS website:
SipHash streaming implementation:
SipHash reference implementation:

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.