hashing {openssl}

Vectorized hash/hmac functions

Description

All hash functions either calculate a hash-digest for key == NULL or HMAC (hashed message authentication code) when key is not NULL. Supported inputs are binary (raw vector), strings (character vector) or a connection object.

Usage

sha1(x, key = NULL)

sha224(x, key = NULL)

sha256(x, key = NULL)

sha384(x, key = NULL)

sha512(x, key = NULL)

keccak(x, size = 256, key = NULL)

sha2(x, size = 256, key = NULL)

sha3(x, size = 256, key = NULL)

md4(x, key = NULL)

md5(x, key = NULL)

blake2b(x, key = NULL)

blake2s(x, key = NULL)

ripemd160(x, key = NULL)

multihash(x, algos = c("md5", "sha1", "sha256", "sha384", "sha512"))

Arguments

x

character vector, raw vector or connection object.

key

string or raw vector used as the key for HMAC hashing

size

must be equal to 224 256 384 or 512

algos

string vector with names of hashing algorithms

Details

The most efficient way to calculate hashes is by using input connections, such as a file() or url() object. In this case the hash is calculated streamingly, using almost no memory or disk space, regardless of the data size. When using a connection input in the multihash function, the data is only read only once while streaming to multiple hash functions simultaneously. Therefore several hashes are calculated simultanously, without the need to store any data or download it multiple times.

Functions are vectorized for the case of character vectors: a vector with n strings returns n hashes. When passing a connection object, the contents will be stream-hashed which minimizes the amount of required memory. This is recommended for hashing files from disk or network.

The sha2 family of algorithms (sha224, sha256, sha384 and sha512) is generally recommended for sensitive information. While sha1 and md5 are usually sufficient for collision-resistant identifiers, they are no longer considered secure for cryptographic purposes.

In applications where hashes should be irreversible (such as names or passwords) it is often recommended to use a random key for HMAC hashing. This prevents attacks where we can lookup hashes of common and/or short strings. See examples. A common special case is adding a random salt to a large number of records to test for uniqueness within the dataset, while simultaneously rendering the results incomparable to other datasets.

The blake2b and blake2s algorithms are only available if your system has libssl 1.1 or newer.

References

Digest types: https://www.openssl.org/docs/man1.1.1/man1/openssl-dgst.html

Examples

# Support both strings and binary
md5(c("foo", "bar"))
md5("foo", key = "secret")

hash <- md5(charToRaw("foo"))
as.character(hash, sep = ":")

# Compare to digest
digest::digest("foo", "md5", serialize = FALSE)

# Other way around
digest::digest(cars, skip = 0)
md5(serialize(cars, NULL))

# Stream-verify from connections (including files)
myfile <- system.file("CITATION")
md5(file(myfile))
md5(file(myfile), key = "secret")

## Not run: check md5 from: http://cran.r-project.org/bin/windows/base/old/3.1.1/md5sum.txt
md5(url("http://cran.r-project.org/bin/windows/base/old/3.1.1/R-3.1.1-win.exe"))
## End(Not run)

# Use a salt to prevent dictionary attacks
sha1("admin") # googleable
sha1("admin", key = "random_salt_value") #not googleable

# Use a random salt to identify duplicates while anonymizing values
sha256("john") # googleable
sha256(c("john", "mary", "john"), key = "random_salt_value")