Password Key Derivation

Password key derivation with Argon2i.

Synopsis

#include <monocypher.h>

void
crypto_argon2i(uint8_t  *hash         , uint32_t hash_size,
               void     *work_area    ,
               uint32_t  nb_blocks    ,
               uint32_t  nb_iterations,
               const uint8_t *password, uint32_t password_size,
               const uint8_t *salt    , uint32_t salt_size);

void
crypto_argon2i_general(
    uint8_t  *hash         ,  uint32_t hash_size,
    void     *work_area    ,
    uint32_t  nb_blocks    ,
    uint32_t  nb_iterations,
    const uint8_t *password,  uint32_t password_size,
    const uint8_t *salt    ,  uint32_t salt_size,
    const uint8_t *key     ,  uint32_t key_size,
    const uint8_t *ad      ,  uint32_t ad_size);

Description

Argon2i is a resource intensive password key derivation scheme optimised for the typical x86-like processor. It runs in constant time with respect to the contents of the password.

Typical applications are password checking (for online services), and key derivation (for encryption). Derived keys can be used to encrypt private keys or password databases.

The version provided by Monocypher has no threading support, so the degree of parallelism is limited to 1. This is considered good enough for most purposes.

The arguments to crypto_argon2i() are:

• hash: The output hash.

• hash_size: Length of hash in bytes. This argument should be set to 32 or 64 for compatibility with the crypto_verify*() constant time comparison functions.

• work_area: Temporary buffer for the algorithm. Must be nb_blocks × 1024 bytes big, and suitably aligned for 64-bit integers. malloc() will yield a suitable chunk of memory.

  The work area is automatically wiped by crypto_argon2i().

• nb_blocks: The number of blocks for the work area. Must be at least 8. A value of 100000 (one hundred megabytes) is a good starting point. If the computation takes too long, reduce this number. If it is too fast, increase this number. If it is still too fast with all available memory, increase nb_iterations.

• nb_iterations: The number of iterations. It must be at least 1. A value of 3 is recommended.

• password: The password to hash. It should be wiped with crypto_wipe() after being hashed.

• password_size: Length of password, in bytes.

• salt: A password salt. This should be filled with random bytes, generated separately for each password to be hashed. See the introduction about random number generation (use your operating system's random number generator).

• salt_size: Length of salt, in bytes. Must be at least 8. 16 is recommended.

The output hash must not overlap with the work area, or it will be wiped along with it. Any other overlap is permitted.

Use crypto_verify16() , crypto_verify32() or crypto_verify64() to compare password hashes to prevent timing attacks.

To select the nb_blocks and nb_iterations parameters, it should first be decided how long the computation should take. For user authentication, we recommend somewhere between half a second (convenient) and several seconds (paranoid). The computation should use as much memory as can be spared.

Since parameter selection depends on your hardware, some trial and error will be required in order to determine the ideal settings. Three iterations and 100000 blocks (that is, one hundred megabytes of memory) is a good starting point. Adjust nb_blocks first. If using all available memory is not slow enough, increase nb_iterations.

crypto_argon2i_general is a variant of crypto_argon2i that supports keyed hashing and hashing of additional data. The additional arguments are:

• key: A key to use in the hash. Can be NULL if key_size is zero. The key is generally not needed, but it does have some uses. In the context of password derivation, it would be stored separately from the password database, and would remain secret even if an attacker were to steal the database. Note that changing the key requires rehashing the user's password, which is only possible upon user login.

• key_size: Length of key, in bytes. Must be zero if there is no key.

• ad: Additional data. This is additional data that goes into the hash, similar to the authenticated encryption with authenticated data (AEAD) construction in crypto_lock_aead() . This most likely has no practical application but is exposed for the sake of completeness. This parameter may be NULL if ad_size is zero.

• ad_size: Length of ad, in bytes. Must be zero if there is no additional data.

Return values

These functions return nothing.

Examples

This example shows how to hash a password with the recommended baseline parameters:

uint8_t        hash[32];                    /* Output hash     */
const uint8_t *password;                    /* User's password */
uint8_t        password_size;               /* Password length */
const uint8_t  salt[16];                    /* Random salt     */
const uint32_t nb_blocks = 100000;          /* 100 megabytes   */
const uint32_t nb_iterations = 3;           /* 3 iterations    */
void *work_area = malloc(nb_blocks * 1024); /* Work area       */
if (work_area == NULL) {
    /* Handle malloc() failure */
}
crypto_argon2i(hash, 32,
               work_area, nb_blocks, nb_iterations,
               password, password_size,
               salt, 16);
/* Wipe secrets if they are no longer needed */
crypto_wipe(password, password_size);
free(work_area);

Standards

These functions implement Argon2i. An RFC draft is being maintained.

Caveats

Any deviation from the specified input and output length ranges results in undefined behaviour. Make sure your inputs are correct.