ISARA Radiate™ Quantum-Safe Library 3.0 Developer’s Guide

Introduction

The ISARA Radiate Quantum-Safe Library gives you the cryptographic building blocks to create applications that will resist attacks by quantum computers.

For more information about ISARA and our quantum-safe solutions, visit www.isara.com.

This Developer’s Guide tells you how to do things with the toolkit, such as:

use Samwise lattice-based key agreement
use the Hierarchical Signature Scheme (HSS) for digital signatures
provide your own implementations of generic algorithms, such as hashes

The toolkit’s API is designed around generic algorithms; you create an instance of the algorithm you want (say, SHA2-256) and then use it with the generic algorithm APIs (iqr_HashBegin(), iqr_HashUpdate(), and iqr_HashEnd()).

Things like signature schemes and key agreement protocols are specialized enough to require customized APIs. We’ve designed these to follow the generic algorithm APIs as much as possible.

We recommend that you have an advanced understanding of cryptography and the theory of security protocols.

The Developer’s Guide covers the following topics:

Getting Started — Foundational information for using the toolkit
Hashing — How to use the toolkit’s SHA2, and SHA3 implementations, and how to provide your own implementations
Random Numbers — How to generate pseudorandom bytes
Message Authentication Codes — How to use the HMAC and Poly1305 MACs
Key Derivation Functions — How to use the KDFs
Symmetric Encryption — How to use the ChaCha20 symmetric-key encryption algorithm
Digital Signatures — How to use the Dilithium, HSS, SPHINCS+, and XMSS digital signature schemes
Key Agreement — How to use the FrodoDH, Samwise, and SIDH key agreement schemes
Key Encapsulation — How to use the Classic McEliece KEM, FrodoKEM, Kyber, NTRUPrime, Saber, and SIKE key encapsulation mechanisms
Technical Info — Detailed information about the compiler options we’ve used, how to maximize code stripping when using the static toolkit, and other details that might be helpful
Building On Windows — How to use the toolkit on the Windows platform

Classical and Quantum Security

Cryptographic algorithms are often described as providing "x bits of security" when used with a certain set of parameters. These parameters can be tuned to provide additional security, usually at the cost of additional CPU and/or memory usage.

These x bits of security apply to an adversary equipped with classical, non-quantum computers, and will appear as "x bits of classical security" in the toolkit’s documentation. The strength of an algorithm against attacks by quantum computers will be written as "x bits of quantum security."

Some classes of algorithm, such as encryption schemes based on the discrete log problem, are thought to be easily breakable by adversaries equipped with quantum computers.

Some algorithms, such as cryptographic hashes and symmetric encryption schemes, are not known or believed to be efficiently breakable by a quantum adversary. Their quantum security strength is generally considered to be half of their classical strength. For example, AES-256 provides 256 bit classical security, and 128 bit quantum security.

Packaging

The toolkit archive contains the following files and directories:

README.html — Information about the toolkit package
SECURITY.html — Information about reporting security issues, and ISARA’s PGP public key
doc — API documentation and this Developer’s Guide document
include — Toolkit headers
lib_<cpu> — Toolkit static and shared libraries tuned for various CPU architectures (for example, lib_skylake would have binaries optimized for the "skylake" architecture)
samples — Sample programs demonstrating how to use the toolkit

Cryptographic signatures for the installation archives are distributed with the archives. If you don’t have access to the signatures, contact support@isara.com.

Getting Help

The latest version of the toolkit documentation is available on the ISARA website. You can also request support via email.

ISARA website
ISARA Radiate Quantum-Safe Library 3.0 Documentation
support@isara.com
1-877-319-8576 Toll-free (Please refer to your support contract.)

Reporting Security Issues

The ISARA team takes security bugs in the toolkit seriously. We appreciate your efforts to responsibly disclose your findings, and will make every effort to acknowledge your contributions.

To report a security issue, email security@isara.com and include the word "SECURITY" in the subject line.

The ISARA team will send a response indicating the next steps in handling your report. After the initial reply to your report, the team will keep you informed of the progress towards a fix and full announcement, and may ask for additional information or guidance.

ISARA’s PGP Key

If you need to validate signatures on files coming from ISARA, or you want to encrypt a file you’re sending to us, you can grab our PGP public key from https://www.isara.com/public_key.asc.

It has the following digests:

Table 1. ISARA PGP Key Digests
Algorithm	Digest
MD5	`9bd6531fee9e6be983a22f0da3415029`
SHA1	`0fac805775c29bb23a769e4c28e36c1c01ec9325`
SHA2-256	`10edbd26d3e470f48ca24e404fa78f45f494ecf492fc6f76ea77a5eb433328d0`
SHA2-512	`726b71a09803801135b02ea4bb87d8bbef9bd9c0f0926ebab6aa7cd2cdaa1375` `140f94d572c32ce1b16a10b771e444397882cb7c425e96cf95e22498e7c30161`

Samples

The samples directory has a number of sub-directories, each with a self-contained program inside demonstrating how to use the toolkit for a specific purpose:

aead_chacha20_poly1305 — Encrypt/decrypt using ChaCha20/Poly1305 for authenticated encryption.
chacha20 — Encrypt/decrypt using ChaCha20.
classicmceliece — Generate keys, encapsulate and decapsulate data using the Classic McEliece KEM.
common — Common functions used by most of the other samples.
dilithium — Generate Dilithium keys, sign a file’s data with a Dilithium key, and verify a Dilithium signature.
frododh — Agree on a shared secret using the FrodoDH scheme.
frodokem — Generate keys, encapsulate and decapsulate data using FrodoKEM.
hash — Hash a file’s data using SHA2-256, SHA2-384, SHA2-512, SHA3-256, or SHA3-512.
hss — Generate keys, sign a file’s data, detach signatures from a private key state, and verify a signature using the HSS algorithm.
integration — Samples showing how to use OpenSSL, /dev/urandom, and the Windows CNG with the toolkit.
kdf — Derive a key (some pseudorandom data) using the specified key derivation function.
kyber — Generate keys, encapsulate and decapsulate data using Kyber.
mac — Generate a message authentication code using the specified MAC algorithm.
ntruprime — Generate keys, encapsulate and decapsulate data using NTRUPrime.
rng — Generate pseudorandom bytes using HMAC-DRBG or SHAKE.
saber — Generate keys, encapsulate, and decapsulate data using the Saber algorithm.
samwise — Agree on a shared secret using Samwise, an optimized version of the FrodoDH scheme.
sidh — Agree on a shared secret using Supersingular Isogeny Diffie-Hellman.
sike — Generate keys, encapsulate and decapsulate data using Supersingular Isogeny Key Encapsulation.
sphincs — Generate keys, sign a file’s data, and verify a signature using the SPHINCS+ algorithm.
version — Display the library’s version information.
VisualStudio — Visual Studio 2017 solution and project files for building the samples.
xmss — Generate keys, sign a file’s data, detach signatures from a private key state, and verify a signature using the XMSS algorithm.

To compile the samples, you will need:

C99 compliant compiler; recent versions of clang or gcc are preferred
cmake 3.7 or newer
GNU make 3.8 or newer (or some other build tool supported by your version of cmake)

Note	Don’t build the samples on macOS using `gcc` 8, they will crash before `main()` due to a problem with `-fstack-protector-all`. Use `clang` to produce Mac binaries.

Set your IQR_TOOLKIT_ROOT environment variable to the directory where you unpacked the installation archive. The build will expect to find include and lib inside the IQR_TOOLKIT_ROOT directory.

Compile all of the samples with:

cd to the samples directory.
mkdir build
cd to the build directory.
cmake ..
make

Getting Started

This section gives you an overview of things you’ll need to know to effectively use the toolkit.

Objects in the toolkit follow a standard cycle:

Create object.
Do things with the object.
Destroy the object.

Some things, like hashes, use a generic API; all hashes use the same Begin(), Update(), End() functions.

Even algorithms that don’t use a generic API will have functions that indicate exactly what they do. For example, HSS (a digital signature scheme) has Sign() and Verify(), while Kyber (a key encapsulation scheme) has Encapsulate() and Decapsulate().

How to Choose an Algorithm

The toolkit has a large number of algorithms and variants, giving you a lot of crypto to choose from. You should pick an appropriate algorithm that meets your use case, and an appropriate variant that meets your requirements and constraints.

This series of questions will help guide you to the algorithm and variant you need.

Hashes, random number generators, MACs, and KDFs are all standard, well-known algorithms with specific well-known uses. The questions only cover quantum-safe algorithms.

What are you doing?

Your specific use case will determine your needs:

Encrypt data using a previously established secret key: you want symmetric encryption.
Securely agree on a secret key: you want key encapsulation or key agreement.
Authenticate data: you want a digital signature.

Encrypting Data

If you’ve already got a secret key safely shared between two systems, use Symmetric Encryption to encrypt data. The toolkit provides ChaCha20 for bulk encryption.

Agreeing on a Secret Key

Before deriving a secret key, you must establish a shared secret. Key Encapsulation mechanisms (KEMs) do this securely.

ISARA recommends using SABER for the best performance. If you require the absolute least amount of bandwidth usage and don’t mind higher CPU utilization, use SIKE.

If your protocol requires a more traditional Diffie-Hellman style key agreement, the toolkit provides several Key Agreement schemes that may be easier to integrate. These are generally non-standard variants of standard KEMs, so they aren’t suitable for situations requiring interoperability.

Similarly, use Samwise for the fastest key agreement, or SIDH for the smallest at the cost of significantly more CPU usage.

Authenticating Data

If you require a standardized signature scheme before the NIST competition declares its standards, use HSS or XMSS. Note that safely implementing these algorithms requires extremely careful state handling.

The NIST competition signature schemes (Dilithium and SPHINCS+) both support an effectively infinite number of signatures and don’t have state that needs to be managed.

For the fastest signing and verification, use Dilithium.

To use a hash-based scheme (similar to HSS and XMSS) without the headache of maintaining state, choose SPHINCS+.

Compiling and Samples

After unpacking the toolkit archive, you can start using it by adding the following command-line arguments to your compiler:

-I/path/to/isara_toolkit/include \
-L/path/to/isara_toolkit/lib \
-liqr_toolkit

You can specify an architecture-specific lib_cpu instead of lib where cpu is one of the available CPUs in your archive. For example, the linux archive includes some of these versions of the library:

lib - 64 bit AMD and Intel CPUs (the default)
lib_core2 - Intel Core 2 CPUs
lib_sandybridge - Intel Sandybridge CPUs
lib_skylake - Intel Skylake CPUs

If you require a version of the library tuned for a specific CPU, please contact ISARA’s sales team.

To build all of the samples at once:

cd /path/to/isara_toolkit/samples
mkdir build
cd build
cmake ..
make

To build a specific sample:

cd /path/to/isara_toolkit/samples/algorithm/sample_name
mkdir build
cd build
cmake ..
make

Note	To build the samples, you’ll also need a recent version of `cmake` (version 3.7 or newer): https://cmake.org/ For most systems, you can use your platform’s normal package tools to install it, but you may need to build an up-to-date version. Binaries are also available on the CMake website.

The Context

To create objects and algorithms in the toolkit, you need a Context. The iqr_Context object keeps track of the registered hashing algorithms as well as additional internal information.

One Context object can be used to create as many other toolkit objects as necessary. Use multiple Context objects if you need to provide several different hash implementations, for example.

The iqr_Context object and its APIs are defined in the iqr_context.h header file. To create one:

#include "iqr_context.h"
...
iqr_Context *context = NULL;
iqr_retval result = iqr_CreateContext(&context);
if (result != IQR_OK) {
    // Examine "result" to see what error has occurred.
}

The toolkit’s Create() functions all take a Context as their first argument.

To properly destroy the context:

#include "iqr_context.h"
...
// Create an iqr_Context object.
// Do crypto.
...
iqr_retval result = iqr_DestroyContext(&context);
if (result != IQR_OK) {
    // Examine "result" to see what error has occurred.
}

Standardized Return Values

All of the APIs in the toolkit return an iqr_retval value, as defined in the iqr_retval.h header.

On success, functions return IQR_OK; if an error occurs, the return value will tell you why the error happened.

Use the iqr_StrError() function (also in iqr_retval.h) to convert an iqr_retval value into an English string. Be sure to write your own error messages if you need to support localization.

Note	Depending on your compiler flags, you must check the return values for toolkit functions.

Registering Hashes

Many algorithms require hash implementations. The Context supplies these, but doesn’t provide a default implementation. This reduces the size of the code in your application by only linking in the algorithms you actually need to use. It also lets you provide your own implementations for additional speed or security.

See the iqr_HashRegisterCallbacks() function in iqr_hash.h.

The toolkit provides implementations of the quantum-safe SHA2 and SHA3 hash algorithms:

IQR_HASH_DEFAULT_SHA2_256 - C implementation of SHA2-256
IQR_HASH_DEFAULT_SHA2_384 - C implementation of SHA2-384
IQR_HASH_DEFAULT_SHA2_512 - C implementation of SHA2-512
IQR_HASH_DEFAULT_SHA3_256 - C implementation of SHA3-256
IQR_HASH_DEFAULT_SHA3_512 - C implementation of SHA3-512

For example, to use the toolkit implementation of SHA2-256:

#include "iqr_context.h"
#include "iqr_hash.h"
...
// Create an iqr_Context object.
...
iqr_retval result = iqr_HashRegisterCallbacks(context, IQR_HASHALGO_SHA2_256,
    &IQR_HASH_DEFAULT_SHA2_256);
if (result != IQR_OK) {
    // Examine "result" to see what error has occurred.
}

After this call to iqr_HashRegisterCallbacks() any toolkit APIs that need a SHA2-256 object can create one using the toolkit’s built-in SHA2-256 implementation.

Registering a Watchdog

Some algorithms, such as HSS key generation, can take a long time to complete. In some cases, you might need to signal a supervising process (or a user) that the code is still working.

In iqr_watchdog.h the toolkit provides a function for registering your own callback function. This function is called periodically by long-running algorithms. (See the API documentation for information about which algorithms call the watchdog callback.)

Your watchdog function should look something like this:

#include "iqr_watchdog.h"
...
iqr_retval my_watchdog_function(void *watchdog_data)
{
    // Notify a supervisor process or provide user feedback.
    ...

    if (some_error_condition) {
        // Return an appropriate error value to abandon the long-running
        // algorithm.
        return IQR_ENOMEM;
    }

    return IQR_OK;
}
...

Then use iqr_WatchdogRegisterCallback() to register it:

#include "iqr_context.h"
#include "iqr_watchdog.h"
...
// Create an iqr_Context object.
...
iqr_retval result = iqr_WatchdogRegisterCallback(context, my_watchdog_function,
    my_data);
if (result != IQR_OK) {
    // Examine "result" to see what error has occurred.
}

After this call to iqr_WatchdogRegisterCallback() any toolkit algorithm using this context that invokes the watchdog will call my_watchdog_function(), passing it the pointer my_data.

If the watchdog function returns an error (anything other than IQR_OK), the calling function will exit without finishing its task.

To remove your watchdog function, call iqr_WatchdogRegisterCallback() with NULL as the function pointer:

#include "iqr_context.h"
#include "iqr_watchdog.h"
...
// Create an iqr_Context object.
...
iqr_retval result = iqr_WatchdogRegisterCallback(context, NULL, NULL);
if (result != IQR_OK) {
    // Examine "result" to see what error has occurred.
}

The data pointer is ignored when the function is set to NULL.

Thread Safety

Objects in the toolkit are self-contained. Any data required to use the object is controlled by the toolkit.

Access to the object is not managed by the toolkit. To use toolkit objects in a multi-threaded environment you’ll have to use the operating system’s mutexes or critical section guards carefully. Your use of the toolkit is as thread-safe as you make it so the toolkit can be as fast as possible in a single-threaded environment.

There are no global variables in the toolkit. There are no data structures with shared, static data.

Anything that can be accessed from several threads needs to be locked in a way that makes sense for the object. For example, with a hash, you need to wrap the entire iqr_HashBegin()/iqr_HashUpdate()/iqr_HashEnd() sequence (or the call to iqr_HashMessage()). For an RNG you might only need to lock iqr_RNGGetBytes() calls (but that gets more complex if you’re using one that needs to be reseeded regularly).

Data Hygiene

Parameter objects (iqr_*Params) in the toolkit never contain any cryptographic material.

All objects in the toolkit have their buffers wiped with 0x00 bytes prior to being deallocated.

Buffers (seed data, hash inputs, etc.) passed to toolkit functions are never modified by the toolkit unless explicitly stated in API documentation.

Version Information

The iqr_version.h header gives you access to the library’s version numbers, plus some identifying information about the build.

IQR_VERSION_MAJOR and IQR_VERSION_MINOR provide the major and minor versions for the toolkit. They can be combined as major.minor to produce a version string.
IQR_VERSION_STRING is a verbose version string for the toolkit.

The iqr_VersionGetBuildTarget() function returns a string representation of the target OS, target CPU architecture, and compiler used to build the toolkit, separated by / characters if you need to parse it into separate fields. Give this information to ISARA’s support team if you discover something that looks like a bug.

Note	The `version` sample is an easy way to get this information for ISARA’s support team.

The iqr_VersionGetBuildHash() function returns a string representation of the library’s build hash and build time stamp, separated by a / character if you need to parse it into separate fields. Give this information to ISARA’s support team if you discover something that looks like a bug.

There’s also an iqr_VersionCheck() function that you can use to ensure that your headers match the version of the library you’re using:

#include "iqr_version.h"
...
iqr_retval result = iqr_VersionCheck(IQR_VERSION_MAJOR, IQR_VERSION_MINOR);
if (result == IQR_OK) {
    // Your headers and library are the same version.
    ...
} else {
    // Your headers and library come from different versions. Your code is
    // not likely to compile and/or link properly.
    ...
}

What Next?

These sections can be read in any order, depending on what you need to do:

To hash some data into a digest, see Hashing.
To generate random bytes, see Random Numbers.
To create message authentication codes from data, see MACs.
To derive secret key data, see KDFs.
To perform symmetric encryption using a shared secret, see Symmetric Encryption.
To digitally sign and verify messages, see Digital Signatures.
To securely agree on a shared secret, see Key Agreement.
To generate a shared secret and encapsulate it for secure transport, see Key Encapsulation.
For detailed technical information about how the library was compiled, see Technical Info.
For information about building toolkit projects with Visual Studio, see Building on Windows.

Hashing

Getting a hash digest for a message is a basic building block of many cryptographic algorithms. The toolkit provides implementations of SHA2 and SHA3.

Registering Hashes

As mentioned in the Getting Started section, you must register an implementation. The toolkit doesn’t automatically associate its own hashes with the Context when you create one. This makes it easier for code stripping to remove unused hash implementations from your application.

#include "iqr_context.h"
#include "iqr_hash.h"
...
// Create iqr_Context.
...
iqr_retval result = iqr_HashRegisterCallbacks(context, algorithm,
    implementation);

The toolkit supports the following algorithms and provides the given implementations:

SHA2-256 (IQR_HASHALGO_SHA2_256) — IQR_HASH_DEFAULT_SHA2_256
SHA2-384 (IQR_HASHALGO_SHA2_384) — IQR_HASH_DEFAULT_SHA2_384
SHA2-512 (IQR_HASHALGO_SHA2_512) — IQR_HASH_DEFAULT_SHA2_512
SHA3-256 (IQR_HASHALGO_SHA3_256) — IQR_HASH_DEFAULT_SHA3_256
SHA3-512 (IQR_HASHALGO_SHA3_512) — IQR_HASH_DEFAULT_SHA3_512

For example, to use the toolkit’s SHA3-512 implementation as the default for all SHA3-512 hashing:

#include "iqr_hash.h"
...
// Create iqr_Context.
...
iqr_retval result = iqr_HashRegisterCallbacks(context, IQR_HASHALGO_SHA3_512,
    &IQR_HASH_DEFAULT_SHA3_512);

Note	The `iqr_Context` object tracks all of the supported SHA2, and SHA3 algorithms; you can register implementations for all of them on a single `iqr_Context`.

When you register a hash implementation, the toolkit runs a known-answer test using the supplied hash callbacks. If this test fails, the registration is ignored and iqr_HashRegisterCallbacks() returns an error.

Using Hashes

After an appropriate hash has been registered, you can use it with the generic hashing API. This API is the same for all hashes.

To create a hash object:

#include "iqr_hash.h"
...
// Create iqr_Context.
// Register hash implementations.
...
iqr_Hash *hash = NULL;
iqr_retval result = iqr_HashCreate(context, algorithm, &hash);
if (result != IQR_OK) {
    // Handle error.
}

To use a hash object:

...
// Call Begin() to initialize the hash.
result = iqr_HashBegin(hash);
if (result != IQR_OK) {
    // Handle error.
}

// Call Update() zero or more times to add data to the hash.
result = iqr_HashUpdate(hash, buffer, buffer_size);
if (result != IQR_OK) {
    // Handle error.
}

// Use iqr_HashGetDigestSize() to get the digest size for this hash.
uint8_t *digest = NULL;
size_t digest_size = 0;
result = iqr_HashGetDigestSize(hash, &digest_size);
if (result != IQR_OK) {
    // Handle error.
}

digest = calloc(1, digest_size);
if (digest == NULL) {
    // Handle out-of-memory.
}

// Call End() to finish the operation and get the digest.
result = iqr_HashEnd(hash, digest, digest_size);

To destroy a hash object:

...
result = iqr_HashDestroy(&hash);
if (result != IQR_OK) {
    // Handle error.
}

There’s also an iqr_HashMessage() function that combines the iqr_HashBegin(), iqr_HashUpdate(), and iqr_HashEnd() process into one call:

...
result = iqr_HashMessage(hash, buffer, buffer_size, digest, digest_size);
if (result != IQR_OK) {
    // Handle error.
}

Writing a Hash Implementation

Because the toolkit doesn’t have a default hash implementation associated with the iqr_Context object, you can use the iqr_HashCallbacks structure (from iqr_hash.h) to provide your own code. This is handy if you have hardware that provides fast hashing, or you want to use another library’s implementation.

The iqr_HashCallbacks structure shows you the signatures of the functions you’ll need to implement:

typedef struct {
    iqr_retval (*initialize)(void **state);

    iqr_retval (*begin)(void *state);

    iqr_retval (*update)(void *state, const uint8_t *buf, size_t buf_size);

    iqr_retval (*end)(void *state, uint8_t *digest, size_t digest_size);

    iqr_retval (*cleanup)(void **state);
} iqr_HashCallbacks;

The initialize() function is passed an empty pointer, where you can store any necessary state. Allocate any memory you need.

#include "iqr_hash.h"
#include "iqr_retval.h"
...
iqr_retval myhash_initialize(void **state)
{
    // Sanity-check inputs.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    if (*state != NULL) {
        return IQR_EINVPTR;
    }

    // Allocate whatever state you need. It's OK to leave it
    // NULL if you don't need to track any state.
    myhash_state *myhash = calloc(1, sizeof(myhash_state));
    if (myhash == NULL) {
        return IQR_ENOMEM;
    }
    ...
    *state = myhash;

    return IQR_OK;
}

Your begin() function is called at the start of a new hashing operation. Initialize your hash code, and get ready to accept data.

#include "iqr_hash.h"
#include "iqr_retval.h"
...
iqr_retval myhash_begin(void *state)
{
    // Sanity-check inputs.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    // Perform any other pre-hashing initialization.
    ...

    return IQR_OK;
}

The update() function is passed the state pointer, a buffer, and the size of the buffer in bytes. Add the buffer’s data to your hash state. This function can be called zero or more times during a hashing operation.

iqr_retval myhash_update(void *state, uint8_t *buf, size_t buf_size)
{
    // Sanity-check input.
    if (buf == NULL && buf_size != 0) {
        return IQR_ENULLPTR;
    }

    if (buf_size == 0) {
        return IQR_OK;
    }

    // Add the data from the buffer to your hash.
    ...

    return IQR_OK;
}

The end() function gets the state pointer, a buffer, and the size of the buffer in bytes. Complete the hashing operation and store its digest in the supplied buffer.

iqr_retval myhash_end(void *state, uint8_t *digest, size_t digest_size)
{
    // Sanity-check input.
    if (state == NULL || digest == NULL) {
        return IQR_ENULLPTR;
    }

    // Make sure there's enough room to store your digest.
    if (digest_size != MYHASH_DIGEST_SIZE) {
        return IQR_EINVBUFSIZE;
    }

    // Finish processing the hash.
    ...

    // Extract the digest and store it in the given buffer.
    ...
}

Finally, the cleanup() function gets a pointer to the state pointer. Wipe and deallocate any state you allocated during initialize() and set the state pointer to NULL.

iqr_retval myhash_cleanup(void **state)
{
    // Clean up and deallocate any state you allocated.
    //
    // You may need to substitute your platform's version of memset_s()
    // (a secure memset() that won't be optimized away by the compiler).
    ...
    memset_s(*state, 0, sizeof(myhash_state));
    free(*state);
    *state = NULL;

    return IQR_OK;
}

Note	The `end()` function is always called when a hash object is destroyed, whether or not `begin()` and `update()` succeeded.

Using OpenSSL’s SHA2-256

For a concrete example of creating your own hash implementation, let’s use OpenSSL’s SHA2-256.

First, we’ll write the hash’s initialize(), begin(), update(), end(), and cleanup() functions using calls to the OpenSSL library:

#include "iqr_hash.h"
#include "iqr_retval.h"

#include <openssl/sha.h>

// OpenSSL APIs return 1 for success.
#define OPENSSL_OK 1

static iqr_retval sha2_256_initialize(void **state)
{
    // Sanity-check inputs.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    if (*state != NULL) {
        return IQR_EINVPTR;
    }

    // Allocate an OpenSSL SHA256_CTX to store the state.
    SHA256_CTX *ctx = NULL;
    ctx = calloc(1, sizeof(*ctx));
    if (ctx == NULL) {
        return IQR_ENOMEM;
    }

    *state = ctx;

    return IQR_OK;
}

static iqr_retval sha2_256_begin(void *state)
{
    // Sanity-check inputs.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    SHA256_CTX *ctx = (SHA256_CTX *)state;

    // Let OpenSSL set up its context.
    int rc = SHA256_Init(ctx);
    if (rc != OPENSSL_OK) {
        return IQR_ENOTINIT;
    }

    return IQR_OK;
}

static iqr_retval sha2_256_update(void *state, const uint8_t *buf,
    size_t buf_size)
{
    // Sanity-check input.
    if (buf == NULL && buf_size != 0) {
        return IQR_ENULLPTR;
    }

    if (buf_size == 0) {
        return IQR_OK;
    }

    // In this case, the state can't be NULL.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    SHA256_CTX *ctx = (SHA256_CTX *)state;

    // Pass the buf pointer into the OpenSSL update function.
    int rc = SHA256_Update(ctx, buf, buf_size);
    if (rc != OPENSSL_OK) {
        // Update failed.
        return IQR_EINVOBJECT;
    }

    return IQR_OK;
}

static iqr_retval sha2_256_end(void *state, uint8_t *digest, size_t digest_size)
{
    // Sanity-check input.
    if (state == NULL || digest == NULL) {
        return IQR_ENULLPTR;
    }

    SHA256_CTX *ctx = (SHA256_CTX *)state;

    // Make sure there's enough room to store your digest.
    if (digest_size != IQR_SHA2_256_DIGEST_SIZE) {
        return IQR_EINVBUFSIZE;
    }

    // Pass the digest pointer into the OpenSSL final function.
    int rc = SHA256_Final(digest, ctx);
    if (rc != OPENSSL_OK) {
        return IQR_EINVOBJECT;
    }

    return IQR_OK;
}

static iqr_retval sha2_256_cleanup(void **state)
{
    SHA256_CTX *ctx = (SHA256_CTX *)*state;

    memset_s(ctx, 0, sizeof(*ctx));
    free(ctx);
    *state = NULL;

    return IQR_OK;
}

Now that we’ve got an implementation, we need to register it so the rest of the toolkit can use it when an algorithm needs a SHA2-256 hash:

// Create the callback structure.
static const iqr_HashCallbacks openssl_sha2_256 = {
    .initialize = sha2_256_initialize,
    .begin = sha2_256_begin,
    .update = sha2_256_update,
    .end = sha2_256_end,
    .cleanup = sha2_256_cleanup
};

// Register the OpenSSL implementation.
result = iqr_HashRegisterCallbacks(context, IQR_HASHALGO_SHA2_256,
    &openssl_sha2_256);
if (result != IQR_OK) {
    // Handle error.
}

After that, any IQR_HASHALGO_SHA2_256 hash object you create with that iqr_Context object will use the OpenSSL implementation:

iqr_Hash *hash = NULL;
iqr_retval result = iqr_HashCreate(context, IQR_HASHALGO_SHA2_256, &hash);
if (result != IQR_OK) {
    // Handle error.
}

Using Microsoft’s SHA2-256

Windows Vista and newer operating systems feature Microsoft’s Cryptography API: Next Generation (CNG). This example demonstrates a SHA2-256 implementation that wraps Microsoft’s CNG functions. For more information, see the Creating a Reusable Hashing Object documentation.

The Windows API can return a number of error codes. In this example, we ignore the NTSTATUS codes and replace them with our own iqr_retval codes. You must handle NTSTATUS codes appropriately for your application.

First, we’ll write the hash’s initialize(), begin(), update(), end(), and cleanup() functions:

#ifndef WIN32_LEAN_AND_MEAN
#define WIN32_LEAN_AND_MEAN  // Exclude rarely-used stuff from Windows headers.
#endif

#include <windows.h>
#include <bcrypt.h>
#include <ntstatus.h>

// This used to be in <ntstatus.h>, but was removed in recent versions of the
// Platform SDK.
#ifndef NT_SUCCESS
#define NT_SUCCESS(status) (status >= 0)
#endif

#include "iqr_hash.h"
#include "iqr_retval.h"

// This will be our state.
typedef struct {
    BCRYPT_ALG_HANDLE alg;
    BCRYPT_HASH_HANDLE hash;
} BCRYPT_HANDLES;

static iqr_retval cng_sha2_256_initialize(void **state)
{
    // Sanity-check inputs.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    if (*state != NULL) {
        return IQR_EINVPTR;
    }

    BCRYPT_HANDLES *handles = NULL;

    // Allocate the state that will be passed around.
    handles = calloc(1, sizeof(BCRYPT_HANDLES));
    if (handles == NULL) {
        return IQR_ENOMEM;
    }

    // A reusable hash object will automatically reset for reuse following a
    // call to BCryptFinishHash(). Thus it is not necessary to recreate a
    // hashing handle.
    DWORD flag = BCRYPT_HASH_REUSABLE_FLAG;

    NTSTATUS status = BCryptOpenAlgorithmProvider(&handles->alg,
        BCRYPT_SHA256_ALGORITHM, NULL, flag);
    if (NT_SUCCESS(status) == false) {
        goto cleanup;
    }

    status = BCryptCreateHash(handles->alg, &handles->hash, NULL, 0, NULL, 0, flag);
    if (NT_ERROR(status)) {
        goto cleanup;
    }

    // Success! Set the state and exit.
    *state = handles;

    return IQR_OK;

cleanup:
    BCryptCloseAlgorithmProvider(handles->alg, 0);
    free(handles);

    return IQR_ENOTINIT;
}

static iqr_retval cng_sha2_256_begin(void *state)
{
    // A reusable hash object automatically resets its internal state.
    //
    // By reusing a hash object, this implementation favours speed by
    // utilizing more memory. An alternative implementation could call
    // BCryptCreateHash() here without the use of BCRYPT_HASH_REUSABLE_FLAG.
    return IQR_OK;
}

static iqr_retval cng_sha2_256_update(void *state, const uint8_t *buf,
    size_t buf_size)
{
    // Sanity-check input.
    if (buf == NULL && buf_size != 0) {
        return IQR_ENULLPTR;
    }

    if (buf_size == 0) {
        return IQR_OK;
    }

    // In this case, the state can't be NULL.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    BCRYPT_HANDLES *handles = (BCRYPT_HANDLES *)state;

    NTSTATUS status = BCryptHashData(handles->hash, (PUCHAR)buf, (ULONG)buf_size, 0);
    if (NT_SUCCESS(status) == false) {
        return IQR_EINVOBJECT;
    }

    return IQR_OK;
}

static iqr_retval cng_sha2_256_end(void *state, uint8_t *digest, size_t digest_size)
{
    // Sanity-check input.
    if (state == NULL || digest == NULL) {
        return IQR_ENULLPTR;
    }

    // Make sure there's enough room to store your digest.
    if (digest_size != IQR_SHA2_256_DIGEST_SIZE) {
        return IQR_EINVBUFSIZE;
    }

    BCRYPT_HANDLES *handles = (BCRYPT_HANDLES *)state;

    // The hash object automatically resets once this call succeeds.
    NTSTATUS status = BCryptFinishHash(handles->hash, (PUCHAR)digest, (ULONG)digest_size, 0);
    if (NT_SUCCESS(status) == false) {
        return IQR_EINVOBJECT;
    }

    return IQR_OK;
}

static iqr_retval cng_sha2_256_cleanup(void **state)
{
    // Sanity-check input.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    BCRYPT_HANDLES *handles = (BCRYPT_HANDLES *)*state;

    NTSTATUS status = BCryptDestroyHash(handles->hash);
    if (NT_SUCCESS(status) == false) {
        return IQR_EINVOBJECT;
    }

    status = BCryptCloseAlgorithmProvider(handles->alg, 0);
    if (NT_SUCCESS(status) == false) {
        return IQR_EINVOBJECT;
    }

    memset_s(handles, 0, sizeof(*handles));
    free(handles);
    *state = NULL;

    return IQR_OK;
}

Now we use iqr_HashRegisterCallbacks() to register this implementation as the SHA2-256 hash:

// Create the callback structure.
const iqr_HashCallbacks cng_sha2_256 = {
    .initialize = cng_sha2_256_initialize,
    .begin = cng_sha2_256_begin,
    .update = cng_sha2_256_update,
    .end = cng_sha2_256_end,
    .cleanup = cng_sha2_256_cleanup
};

// Register the CNG implementation.
result = iqr_HashRegisterCallbacks(context, IQR_HASHALGO_SHA2_256,
    &cng_sha2_256);
if (result != IQR_OK) {
    // Handle error.
}

Now, any IQR_HASHALGO_SHA2_256 hash object you create will use the CNG SHA2-2 256 implementation:

iqr_Hash *hash = NULL;
iqr_retval result = iqr_HashCreate(context, IQR_HASHALGO_SHA2_256, &hash);
if (result != IQR_OK) {
    // Handle error.
}

Random Numbers

Generating random data is an important part of many cryptographic algorithms. The toolkit supports the HMAC-DRBG algorithm for generating data.

For simplicity, we refer to this class of algorithm as random number generators (RNGs).

Seed Data

Pseudo-random number generators are only as good as the seed data you use to initialize them. This seed data must come from a good source of entropy.

Refer to your target system’s CPU or OS documentation to find the best source of entropy available to you.

Using a poor source of entropy data will compromise the randomness of the data produced by these algorithms.

Using RNGs

Like hashes, RNGs use a generic API. Unlike hashes, the RNGs supported by the toolkit require custom Create() functions to provide suitable initialization data for each algorithm.

To create an HMAC-DRBG object using any of the IQR_HASHALGO_* constants from iqr_hash.h:

#include "iqr_context.h"
#include "iqr_hash.h"
#include "iqr_rng.h"
...
// Create iqr_Context.
// Register a hash implementation for the implementation you want to use
// (IQR_HASHALGO_SHA2_256 in this example).
...
iqr_RNG *rng = NULL;
iqr_retval result = iqr_RNGCreateHMACDRBG(context, IQR_HASHALGO_SHA2_256, &rng);
if (result != IQR_OK) {
    // Handle error.
}

Note	To provide a nonce for HMAC-DRBG, include it in the seed data given to `iqr_RNGInitialize()`. If your seed data was "password" and your nonce is "random", set the initialization buffer to "passwordrandom".

To create an RNG object using your own implementation:

#include "iqr_context.h"
#include "iqr_rng.h"
...
// Create iqr_Context.
...
iqr_RNGCallbacks rng_callbacks = {
    .initialize = my_rng_initialize,
    .reseed = my_rng_reseed,
    .getbytes = my_rng_getbytes,
    .cleanup = my_rng_cleanup
};
...
iqr_RNG *rng = NULL;
iqr_retval result = iqr_RNGCreate(context, &rng_callbacks, &rng);
if (result != IQR_OK) {
    // Handle error.
}

To use an RNG object:

...
// Call Initialize() to initialize the RNG.
result = iqr_RNGInitialize(rng, seed_buffer, seed_size);
if (result != IQR_OK) {
    // Handle error.
}

// Call GetBytes() to get bytes from the RNG.
result = iqr_RNGGetBytes(rng, buffer, buffer_size);
if (result == IQR_ERESEED) {
    // The RNG is depleted, call Reseed() to give it more seed data.
    result = iqr_RNGReseed(rng, new_seed_buffer, new_seed_size);
    if (result != IQR_OK) {
        // Handle error.
    }

    result = iqr_RNGGetBytes(rng, buffer, buffer_size);
    if (result != IQR_OK) {
        // Handle error.
    }
} else if (result != IQR_OK) {
    // Handle error.
}

// You can also reseed the RNG as necessary; you don't
// need to wait until GetBytes() returns IQR_ERESEED.
result = iqr_RNGReseed(rng, new_seed_buffer, new_seed_size);
if (result != IQR_OK) {
    // Handle error.
}

To destroy an RNG object:

...
result = iqr_RNGDestroy(&rng);
if (result != IQR_OK) {
    // Handle error.
}

Writing an RNG Implementation

You can use the iqr_RNGCallbacks structure (from iqr_rng.h) to provide your own RNG implementation. This is handy if you have hardware that provides cryptographically secure random number generation, or you want to use another library’s implementation.

The iqr_RNGCallbacks structure shows you the signatures of the functions you’ll need to implement:

typedef struct {
    iqr_retval (*initialize)(void **state, const uint8_t *seed,
        size_t seed_size);

    iqr_retval (*reseed)(void *state, const uint8_t *entropy,
        size_t entropy_size);

    iqr_retval (*getbytes)(void *state, uint8_t *buf, size_t buf_size);

    iqr_retval (*cleanup)(void **state);
} iqr_RNGCallbacks;

Your initialize() function is passed an empty pointer, which it can use to store any necessary state. In addition, you’re passed a buffer containing seed data, and the size (in bytes) of that buffer. Allocate any memory you need, initialize your RNG code, and get ready to generate bytes.

#include "iqr_retval.h"
#include "iqr_rng.h"
...
iqr_retval myrng_initialize(void **state, const uint8_t *seed, size_t seed_size)
{
    // Sanity-check inputs.
    if (state == NULL || seed == NULL) {
        return IQR_ENULLPTR;
    }

    if (*state != NULL) {
        return IQR_EINVPTR;
    }

    // The caller must provide seed data.
    if (seed_size == 0) {
        return IQR_EINVBUFSIZE;
    }

    // Allocate whatever state you need. It's OK to leave it
    // NULL if you don't need to track any state.
    myrng_state *myrng = calloc(1, sizeof(myrng_state));
    if (myrng == NULL) {
        return IQR_ENOMEM;
    }
    ...
    *state = myrng;

    // Perform any other initialization.
    ...

    return IQR_OK;
}

The reseed() function is passed the state pointer, a buffer, and the size of the buffer in bytes. Add the buffer’s data to your RNG’s internal entropy.

iqr_retval myrng_reseed(void *state, uint8_t *entropy, size_t entropy_size)
{
    // Sanity-check input.
    if (entropy == NULL) {
        return IQR_ENULLPTR;
    }

    if (entropy_size == 0) {
        return IQR_EINVBUFSIZE;
    }

    // Add the data from the additional entropy buffer to your RNG's entropy.
    ...

    return IQR_OK;
}

The getbytes() function is passed the state pointer, a buffer, and the size of the buffer in bytes. Write that many random bytes into the buffer.

iqr_retval myrng_getbytes(void *state, uint8_t *buf, size_t buf_size)
{
    // Sanity-check input.
    if (buf == NULL) {
        return IQR_ENULLPTR;
    }

    if (buf_size == 0) {
        return IQR_EINVBUFSIZE;
    }

    // Generate random bytes and write them into the buffer.
    ...

    return IQR_OK;
}

Finally, the cleanup() function gets a pointer to the state pointer. Before returning, deallocate any state you allocated during initialize() and set the state pointer to NULL.

iqr_retval myrng_cleanup(void **state)
{
    // Sanity-check input.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    // Clean up and deallocate any state you allocated.
    //
    // You may need to substitute your platform's version of memset_s()
    // (a secure memset() that won't be optimized away by the compiler).
    ...
    memset_s(*state, 0, sizeof(myrng_state));
    free(*state);
    *state = NULL;

    return IQR_OK;
}

Note	The `cleanup()` function is always called when an RNG object is destroyed, whether or not `initialize()`, `getbytes()`, and `reseed()` succeeded.

Using /dev/urandom as an RNG

Here’s a concrete example of how to create your own RNG implementation, using /dev/urandom as a source of random bytes. Refer to your operating system’s /dev/urandom documentation for details about its behaviour, and its suitability as a cryptographic random number generator.

This sample assumes your /dev/urandom implementation lets you write additional entropy to the device.

First, we’ll write the RNG’s initialize(), reseed(), getbytes(), and cleanup() functions using POSIX functions and /dev/urandom:

#include <fcntl.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

#include "iqr_context.h"
#include "iqr_retval.h"
#include "iqr_rng.h"

static iqr_retval devurandom_initialize(void **state, const uint8_t *seed,
    size_t seed_size)
{
    // Sanity-check inputs.
    if (state == NULL || seed == NULL) {
        return IQR_ENULLPTR;
    }

    if (*state != NULL) {
        return IQR_EINVPTR;
    }

    // The caller must provide seed data.
    if (seed_size == 0) {
        return IQR_EINVBUFSIZE;
    }

    int *device_handle = calloc(1, sizeof(int));
    if (device_handle == NULL) {
        return IQR_ENOMEM;
    }

    iqr_retval result = IQR_OK;

    // Write our seed data to the device.
    *device_handle = open("/dev/random", O_RDWR);
    if (*device_handle == -1) {
        result = IQR_ENOTINIT;
        goto cleanup;
    }

    ssize_t bytes_written = 0;
    while (seed_size > 0) {
        bytes_written = write(*device_handle, seed, seed_size);
        if (bytes_written == -1) {
            result = IQR_EINVOBJECT;
            goto cleanup;
        }
        seed_size -= (size_t)bytes_written;
    }

    // We don't need to allocate state, just store the file descriptor.
    *state = device_handle;
    return result;

cleanup:
    if (*device_handle != -1) {
        close(*device_handle);
    }
    free(device_handle);
    device_handle = NULL;

    return result;
}

static iqr_retval devurandom_reseed(void *state, const uint8_t *entropy,
    size_t entropy_size)
{
    // Sanity-check input.
    if (state == NULL || entropy == NULL) {
        return IQR_ENULLPTR;
    }

    if (entropy_size == 0) {
        return IQR_EINVBUFSIZE;
    }

    // Add the data to your RNG's entropy.
    int *device_handle = state;
    ssize_t bytes_written = 0;
    while (entropy_size > 0) {
        bytes_written = write(*device_handle, entropy, entropy_size);
        if (bytes_written == -1) {
            return IQR_EINVOBJECT;
        }
        entropy += bytes_written;
        entropy_size -= (size_t)bytes_written;
    }

    return IQR_OK;
}

static iqr_retval devurandom_getbytes(void *state, uint8_t *buf,
    size_t buf_size)
{
    // Sanity-check input.
    if (state == NULL || buf == NULL) {
        return IQR_ENULLPTR;
    }

    if (buf_size == 0) {
        return IQR_EINVBUFSIZE;
    }

    // Generate random bytes and write them into the buffer.
    int *device_handle = state;
    ssize_t bytes_read = 0;
    while (buf_size > 0) {
        bytes_read = read(*device_handle, buf, buf_size);
        if (bytes_read == -1) {
            return IQR_EINVDATA;
        }
        buf += bytes_read;
        buf_size -= (size_t)bytes_read;
    }

    return IQR_OK;
}

static iqr_retval devurandom_cleanup(void **state)
{
    // Sanity-check input.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    // Clean up and deallocate any state you allocated.
    int *device_handle = *state;
    close(*device_handle);
    free(device_handle);
    *state = NULL;

    return IQR_OK;
}

Now that you’ve got an implementation, you can use it with iqr_RNGCreate():

// Create the callback structure.
static const iqr_RNGCallbacks devurandom_rng = {
    .initialize = devurandom_initialize,
    .reseed = devurandom_reseed,
    .getbytes = devurandom_getbytes,
    .cleanup = devurandom_cleanup
};

iqr_RNG *rng = NULL;
iqr_retval result = iqr_RNGCreate(context, &devurandom_rng, &rng);
if (result != IQR_OK) {
    // Handle error.
}

Using Microsoft’s Cryptographic Service as an RNG

Windows Vista and newer operating systems feature Microsoft’s Cryptography API: Next Generation (CNG). Like the previous example, the following code demonstrates a random number generator (RNG) implementation that wraps Microsoft’s random number generation functions. For more information, see the CNG documentation.

Windows constantly adds entropy, so the CNG RNG does not need to be reseeded.

First, we’ll write the RNG’s initialize(), reseed(), getbytes(), and cleanup() functions:

#ifndef WIN32_LEAN_AND_MEAN
#define WIN32_LEAN_AND_MEAN  // Exclude rarely-used stuff from Windows headers.
#endif

#include <windows.h>
#include <bcrypt.h>
#include <ntstatus.h>

#include <stdbool.h>
#include <stdio.h>

// This may not exist in <ntstatus.h>
#ifndef NT_SUCCESS
#define NT_SUCCESS(status) (status >= 0)
#endif

#include "iqr_context.h"
#include "iqr_retval.h"
#include "iqr_rng.h"

static iqr_retval cng_rng_initialize(void **state, const uint8_t *seed,
    size_t seed_size)
{
    (void)seed;  // Unused by Windows CNG.
    (void)seed_size;

    // Sanity-check inputs.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    if (*state != NULL) {
        return IQR_EINVPTR;
    }

    BCRYPT_ALG_HANDLE alg = NULL;

    // Open a handle to the CNG.
    NTSTATUS status = BCryptOpenAlgorithmProvider(&alg, BCRYPT_RNG_ALGORITHM, NULL, 0);
    if (NT_SUCCESS(status) == false) {
        return IQR_ENOTINIT;
    }

    *state = alg;
    return IQR_OK;
}

static iqr_retval cng_rng_reseed(void *state, const uint8_t *entropy,
    size_t entropy_size)
{
    (void)state;  // Unused variables.
    (void)entropy;
    (void)entropy_size;

    // The CNG random number generator will never need reseeding.
    return IQR_OK;
}

static iqr_retval cng_rng_getbytes(void *state, uint8_t *buf, size_t buf_size)
{
    // Sanity-check input.
    if (state == NULL || buf == NULL) {
        return IQR_ENULLPTR;
    }

    if (buf_size == 0) {
        return IQR_EINVBUFSIZE;
    }

    BCRYPT_ALG_HANDLE alg = (BCRYPT_ALG_HANDLE)state;

    // Generate requested random bytes.
    NTSTATUS status = BCryptGenRandom(alg, (PUCHAR)buf, (ULONG)buf_size, 0);
    if (NT_SUCCESS(status) == false) {
        return IQR_EINVOBJECT;
    }

    return IQR_OK;
}

static iqr_retval cng_rng_cleanup(void **state)
{
    // Sanity-check input.
    if (state == NULL) {
        return IQR_ENULLPTR;
    }

    BCRYPT_ALG_HANDLE alg = (BCRYPT_ALG_HANDLE)*state;

    // Close the provider handle.
    NTSTATUS status = BCryptCloseAlgorithmProvider(alg, 0);
    if (NT_SUCCESS(status) == false) {
        return IQR_EINVOBJECT;
    }

    *state = NULL;
    return IQR_OK;
}

Once the implementation is finished, it’s ready to use with iqr_RNGCreate():

// Create the callback structure.
static const iqr_RNGCallbacks cng_rng = {
    .initialize = cng_rng_initialize,
    .reseed = cng_rng_reseed,
    .getbytes = cng_rng_getbytes,
    .cleanup = cng_rng_cleanup
};

iqr_RNG *rng = NULL;
iqr_retval result = iqr_RNGCreate(context, &cng_rng, &rng);
if (result != IQR_OK) {
    // Handle error.
}

Message Authentication Codes

The toolkit provides two message authentication code algorithms, HMAC and Poly1305.

MACs have the following key length requirements, based on the algorithm:

HMAC (SHA2-256 or SHA3-256) - 32 bytes or more
HMAC (SHA2-512 or SHA3-512) - 64 bytes or more
Poly1305 - 32 bytes (More is allowed, but additional data will be ignored.)

Using MACs

Using a MAC is similar to using a hash:

Create the MAC using iqr_MACCreateHMAC() or iqr_MACCreatePoly1305().
Begin the MAC operation.
Update it with data.
End the MAC and get the tag.
Destroy the MAC, or begin again.

The hash-based message authentication code (HMAC) algorithm (found in iqr_mac.h) requires you to register an implementation for the hash algorithm you want to use; see the hashing section for more information about hash algorithms.

To create an HMAC:

#include "iqr_hash.h"
#include "iqr_mac.h"
...
// Create iqr_Context.
// Register hash implementations.
...
iqr_MAC *mac = NULL;
iqr_retval result = iqr_MACCreateHMAC(context, algorithm, &mac);
if (result != IQR_OK) {
    // Handle error.
}

The Poly1305 message authentication code algorithm (found in iqr_mac.h) doesn’t have any external dependencies, but does require a one-time key. It can be combined with the ChaCha20 cipher to provide Authenticated Encryption with Associated Data (AEAD) as demonstrated in the aead_chacha20_poly1305 sample (in your samples directory, or found on GitHub).

To create a Poly1305 MAC:

#include "iqr_mac.h"
...
// Create iqr_Context.
...
iqr_MAC *mac = NULL;
iqr_retval result = iqr_MACCreatePoly1305(context, &mac);
if (result != IQR_OK) {
    // Handle error.
}

To use a MAC object:

...
// Call Begin() to initialize the MAC.
result = iqr_MACBegin(mac, key, key_size);
if (result != IQR_OK) {
    // Handle error.
}

// Call Update() zero or more times to add data to the MAC.
result = iqr_MACUpdate(mac, buffer, buffer_size);
if (result != IQR_OK) {
    // Handle error.
}

// Use iqr_MACGetTagSize() to get the tag size for this MAC.
uint8_t *tag = NULL;
size_t tag_size = 0;
result = iqr_MACGetTagSize(mac, &tag_size);
if (result != IQR_OK) {
    // Handle error.
}

tag = calloc(1, tag_size);
if (tag == NULL) {
    // Handle out-of-memory.
}

// Call End() to finish the operation and get the tag.
result = iqr_MACEnd(mac, tag, tag_size);
if (result != IQR_OK) {
    // Handle error.
}

To destroy a MAC object:

...
result = iqr_MACDestroy(&mac);
if (result != IQR_OK) {
    // Handle error.
}

There’s also an iqr_MACMessage() function that combines the iqr_MACBegin(), iqr_MACUpdate(), iqr_MACEnd() process into one call:

...
result = iqr_MACMessage(mac, key, key_size, buffer, buffer_size, tag, tag_size);
if (result != IQR_OK) {
    // Handle error.
}

Key Derivation Functions

The toolkit provides three standard key derivation functions (KDFs):

RFC 5869’s HMAC-based extract-and-expand KDF (HKDF)
NIST SP 800-56C Option 1 Concatenation KDF
RFC 2898’s Password Based Key Derivation Function 2 (PBKDF2)

Because the KDFs all have slightly different needs, there is currently no generic KDF API.

RFC 5869 HKDF

Because the RFC 5869 KDF uses an HMAC internally, you must register a hashing algorithm before using the KDF.

To use the RFC 5869 HMAC-based KDF:

#include "iqr_context.h"
#include "iqr_hash.h"
#include "iqr_kdf.h"
...
// Create iqr_Context.
// Register hashing algorithms.
...
iqr_retval result = iqr_RFC5869HKDFDeriveKey(context, hash_algorithm,
    salt_buffer, salt_size, ikm_buffer, ikm_size, info_buffer, info_size,
    key_buffer, key_size);
if (result != IQR_OK) {
    // Handle error.
}

Use the salt buffer to provide additional randomness. The salt buffer can be NULL (and the salt buffer size 0), but providing salt will improve the security of your application.

The initial keying material (the IKM buffer) is similar to the seeding data given to a random number generator. Some algorithms may have an existing cryptographically strong key to use for the initial keying material, such as the premaster secret in TLS RSA cipher suites. You must provide data in this buffer.

The optional info buffer is for context and application specific information. This binds the derived key to your information, such as a protocol number, an algorithm identifier, user data, etc.

The derived key data is returned in the key buffer. The key size cannot be more than 254 times the size of the hash’s digest size or this will return an IQR_EOUTOFRANGE error.

Note	RFC 5869 HKDF is the cryptographically strongest KDF currently provided by the toolkit.

NIST Concatenation KDF

Because the NIST SP 800-56C Option 1 Concatenation KDF uses a hash internally, you must register a hashing algorithm before using the KDF.

To use the Concatenation KDF:

#include "iqr_context.h"
#include "iqr_hash.h"
#include "iqr_kdf.h"
...
// Create iqr_Context.
// Register hashing algorithms.
...
iqr_retval result = iqr_ConcatenationKDFDeriveKey(context, hash_algorithm,
    shared_secret, shared_secret_size, other_info, other_info_size,
    key_buffer, key_size);
if (result != IQR_OK) {
    // Handle error.
}

The shared secret ("Z" in the specification) is pre-determined data shared between all systems generating keys with the Concatenation KDF. The shared secret must be at least one byte.

The other info ("OtherInfo" in the specification) is for context and application specific information. This binds the derived key to your information, such as a protocol number, an algorithm identifier, user data, etc. It can be NULL (and set the buffer size to 0).

The derived key is returned in the key buffer.

RFC 2898 PBKDF2

Because the RFC 2898 PBKDF2 uses a hash internally, you must register a hashing algorithm before using the KDF.

To use PBKDF2:

#include "iqr_context.h"
#include "iqr_hash.h"
#include "iqr_kdf.h"
...
// Create iqr_Context.
// Register hashing algorithms.
...
iqr_retval result = iqr_PBKDF2DeriveKey(context, hash_algorithm,
    password, password_size, salt, salt_size, iteration_count,
    key_buffer, key_size);
if (result != IQR_OK) {
    // Handle error.
}

The optional password is pre-determined data shared between all systems generating keys with PBKDF2. This can be NULL if the password size is also 0 bytes.

Use the salt buffer to provide additional randomness. The salt buffer can be NULL (and the salt buffer size 0), but providing salt will improve the security of your application.

Using both a password and a salt provides the best security.

PBKDF2 uses the specified number of iterations to improve the randomness of its derived data at the expense of additional processing time. Use the maximum value that’s tolerable for your application or the value specified by your protocol.

The derived key is returned in the key buffer.

Symmetric Encryption

In general, symmetric encryption schemes are not significantly threatened by quantum computers. Doubling the key sizes provides enough security in the face of a quantum threat.

The toolkit currently provides one symmetric algorithm, RFC 8439’s ChaCha20.

ChaCha20

ChaCha20 (see iqr_chacha20.h) is an easy to use cipher that doesn’t require additional parameter objects. The key and nonce are provided as byte buffers.

To encrypt data using ChaCha20:

#include "iqr_chacha20.h"
#include "iqr_context.h"
...
// Create iqr_Context. Not strictly needed for ChaCha20, but needed for
// other toolkit APIs.
...
iqr_retval result = iqr_ChaCha20Encrypt(key_buffer, key_size, nonce, nonce_size,
    counter, plaintext, plaintext_size, ciphertext, ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

The key buffer must have exactly IQR_CHACHA20_KEY_SIZE bytes of data, and the nonce buffer must have exactly IQR_CHACHA20_NONCE_SIZE bytes of data. The key can be pre-shared using a suitable key agreement protocol, and the nonce should be unique per encryption stream.

Use the counter to indicate the start of this block. Because ChaCha20 is a block cipher operating in counter mode, you must increment the counter by the number of encrypted blocks (that is, ceiling(plaintext_size / 64)) when encrypting additional data using the same key and nonce.

The ciphertext buffer must be at least as large as the plaintext buffer.

To decrypt data using ChaCha20:

...
result = iqr_ChaCha20Decrypt(key_buffer, key_size, nonce, nonce_size,
    counter, ciphertext, ciphertext_size, plaintext, plaintext_size);
if (result != IQR_OK) {
    // Handle error.
}

The key, nonce, and counter must match the values used to encrypt the data.

Since ChaCha20 is a symmetric cipher, encrypt and decrypt are the same operation, with plaintext and ciphertext swapped.

Digital Signatures

The toolkit provides these digital signature schemes:

Dilithium — a lattice-based signature scheme
HSS (Hierarchical Signature Scheme) — a hash-based one-time signature scheme
SPHINCS+ — a stateless hash-based signature scheme
XMSS (eXtended Merkle Signature Scheme) — a hash-based one-time signature scheme

If the message you’re signing is a digest from a hash function, it should be 32 bytes or longer to ensure that the signature is quantum-safe. Use 48 bytes or more if you have high security requirements.

Digital Signature Usage

Alice creates a key pair.
- Alice makes her public key available to anyone.
- Alice keeps her private key secret.
Alice uses her private key to sign a message.
- The message can be any data, such as a hash digest, the contents of a file, or a buffer.
- Alice distributes the signature along with the message.
Anyone can use her public key, the message, and the signature, to verify that the message was signed by the private key that matches the public key.

HSS and XMSS Tree Strategies

Tree strategies offer a trade-off between CPU utilization and memory usage during signing. Choosing the correct strategy is highly dependent on the hardware restrictions of the target platform.

MEMORY_CONSTRAINED_STRATEGY: The Memory Constrained strategy implements an algorithm that minimizes the memory/storage requirements of the state at the cost of recomputing parts of the tree during signing. This option is ideal for memory constrained devices with a fast CPU to handle the extra computation.
FULL_TREE_STRATEGY: The Full Tree strategy keeps the entire Merkle tree in memory. This strategy uses the least CPU at the cost of using memory. This option is ideal for servers with large amounts of memory and the need to generate signatures frequently.
VERIFY_ONLY_STRATEGY: The Verify strategy is only used to verify signatures; it cannot be used to create or import private keys nor can it be used to create signatures. This option is ideal for a client that only needs to verify signatures. It uses the least CPU and RAM.

Note	Using the `iqr_HSSCreateParamsFromSignature()` function (which automatically sets the `VERIFY_ONLY_STRATEGY`), you can verify signatures created using any valid combination of HSS parameters.

HSS and XMSS States

The HSS and XMSS algorithms are stateful hash-based signatures. Their private key must be accompanied by a state that gets updated every time you perform a Sign() operation to generate a signature.

The size (and contents) of the state depend on height and the tree strategy chosen when generating keys. See the Technical Information section for more information.

Each time the Sign() function is called, the state is advanced to the next usable state. You must store this new state in non-volatile memory prior to releasing the signature. Alternatively, use DetachState() to work with smaller states that can be lost without catastrophe.

Use the DetachState() function to split a state into two distinct, non-overlapping state objects. This effectively "reserves" a number of signatures from the state and places them into a new detached state. Both state objects must be used with the same private key to generate signatures.

This operation is useful for disaster recovery. You can detach a small section of the state and use it for signing while the rest of the state is persisted in non-volatile memory. In the event of a power outage, the persisted state will not issue signatures that overlap with the state lost in volatile memory.

Important

Using DetachState() to move private data from one signing device to another violates the constraints defined by NIST SP 800-208. The resulting detached state must not leave the signing hardware.

The detached state will have a smaller number of maximum signatures once DetachState() returns. Use GetSignatureCount() to obtain the number of available signatures. Use GetStateSize() to obtain the state sizes prior to exporting.

Dilithium Signature Scheme

The toolkit provides an implementation of the Dilithium signature scheme as defined in CRYSTALS - Dilithium. The toolkit implements the SHAKE-based variants of Dilithium.

Dilithium comes in three variants:

IQR_DILITHIUM_2 — NIST security level II
IQR_DILITHIUM_3 — NIST security level III
IQR_DILITHIUM_5 — NIST security level V

No additional parameters are necessary.

Creating Keys

The toolkit lets you create Dilithium keys after choosing the variant.

To create a Dilithium key pair:

#include "iqr_context.h"
#include "iqr_dilithium.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// Create Dilithium parameters using one of the variants from iqr_dilithium.h.
iqr_DilithiumParams *params = NULL;
iqr_retval result = iqr_DilithiumCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

// Create the key pair.
iqr_DilithiumPublicKey *public_key = NULL;
iqr_DilithiumPrivateKey *private_key = NULL;
result = iqr_DilithiumCreateKeyPair(params, rng, &public_key, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

Signing

To sign a message using the Dilithium private key:

...
size_t signature_size = 0;
result = iqr_DilithiumGetSignatureSize(params, &signature_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *signature = calloc(1, signature_size);
if (signature == NULL) {
    // Handle error.
}

result = iqr_DilithiumSign(private_key, digest, digest_size, signature,
    signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Verifying Signatures

To verify a signature using the Dilithium public key:

...
result = iqr_DilithiumVerify(public_key, digest, digest_size, signature,
    signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Managing Keys

To export Dilithium keys for storage or transmission:

...
size_t public_key_data_size = 0;
result = iqr_DilithiumGetPublicKeySize(params, &public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_data = calloc(1, public_key_data_size);
if (public_key_data == NULL) {
    // Handle error.
}

result = iqr_DilithiumExportPublicKey(public_key, public_key_data,
    public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

size_t private_key_data_size = 0;
result = iqr_DilithiumGetPrivateKeySize(params, &private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *private_key_data = calloc(1, private_key_data_size);
if (private_key_data == NULL) {
    // Handle error.
}

result = iqr_DilithiumExportPrivateKey(private_key, private_key_data,
    private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

To import Dilithium keys from buffers:

...
// Create an iqr_DilithiumParams object using the same variant that was used to
// create the keys.
...

iqr_DilithiumPrivateKey *private_key = NULL;
result = iqr_DilithiumImportPrivateKey(params, private_key_data,
    private_key_data_size, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

iqr_DilithiumPublicKey *public_key = NULL;
result = iqr_DilithiumImportPublicKey(params, public_key_data,
    public_key_data_size, &public_key);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for importing keys must match the `params` used to create the keys.

Public Key Format

The data produced by iqr_DilithiumExportPublicKey() follows the format documented in the CRYSTALS - Dilithium document:

uint8_t seed[32];  // The RNG seed used to create the public key.
uint8_t t1[];  // POLY16 encoded t1 data.

Signature Format

The signature produced by iqr_DilithiumSign() follows the format documented in section 3 of the CRYSTALS - Dilithium document:

uint8_t c[];
uint8_t z[];
uint8_t h[];

HSS (Hierarchical Signature Scheme)

The toolkit provides an implementation of the Hierarchical Signature Scheme (HSS) as defined in RFC 8554.

HSS is a signature scheme that has several major differences from classical digital signature schemes:

An HSS private key can only be used to sign a finite number of items.
You need to carefully maintain the state.

The variant can be one of:

IQR_HSS_2E15_FAST — 2¹⁵ (32,768) one-time signatures, trading speed for larger signatures
IQR_HSS_2E15_SMALL — 2¹⁵ (32,768) one-time signatures, reducing speed for smaller signatures

These variants correspond to the following parameters from the RFC:

Table 2. HSS Variant Parameters
Variant	Layers	Height	Winternitz
`IQR_HSS_2E15_FAST`	1	15	2
`IQR_HSS_2E15_SMALL`	1	15	8

It’s up to the user to manage domain parameters.

The size of the HSS state varies depending on the variant and tree strategy used; see the Technical Information section for key and state sizes. HSS signature size is relative to the full tree height and Winternitz value.

Creating Keys

The toolkit lets you create HSS keys by specifying individual parameters.

To create an HSS key pair:

#include "iqr_context.h"
#include "iqr_hash.h"
#include "iqr_hss.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Register a SHA2-256 hash algorithm.
// Create and initialize a Random Number Generator, rng.
...
// Create HSS parameters.
iqr_HSSParams *params = NULL;
iqr_retval result = iqr_HSSCreateParams(context,
    &IQR_HSS_MEMORY_CONSTRAINED_STRATEGY, IQR_HSS_2E15_FAST, &params);
if (result != IQR_OK) {
    // Handle error.
}

// Create the key pair.
iqr_HSSPublicKey *public_key = NULL;
iqr_HSSPrivateKey *private_key = NULL;
iqr_HSSPrivateKeyState *state = NULL;
result = iqr_HSSCreateKeyPair(params, rng, &public_key, &private_key,
    &state);
if (result != IQR_OK) {
    // Handle error.
}

Detaching State

In some cases, you may need to split a key’s state. For disaster recovery, you can reserve some one-time signatures and store the remaining state securely in non-volatile memory. If you lose power, the reserved signatures are lost, but the remaining state can be recovered.

Important

If this feature is used to move private data between two signing devices, it violates the conditions defined in NIST SP 800-208.

This is accomplished by detaching the key state. Using iqr_HSSDetachState() will create a new state containing the number of signatures requested, and remove those signatures from the original state.

To detach num_sigs one-time signatures from an HSS key’s state:

#include "iqr_hss.h"
...
// This assumes you've already got an HSS private key and state, and that
// you've registered a SHA2-256 hash implementation.
...
iqr_HSSPrivateKeyState *new_state = NULL;
iqr_retval result = iqr_HSSDetachState(private_key, original_state,
    num_sigs, &new_state)
if (result != IQR_OK) {
    // Handle error.
}

After calling iqr_HSSDetachState(), num_sigs one-time signatures will have been removed from original_state. The new_state will have the removed signatures.

Signing

To sign a message using the HSS private key:

...
size_t signature_size = 0;
result = iqr_HSSGetSignatureSize(params, &signature_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *signature = calloc(1, signature_size);
if (signature == NULL) {
    // Handle error.
}

/*********************** CRITICALLY IMPORTANT STEP *************************

  You must detach a state to use for signing, and write the remaining state
  to non-volatile memory, before signing. Failure to do so could result in
  a security breach as it could lead to the re-use of a one-time signature.

  This step has been omitted for brevity.

  For more information about this property of the HSS state, please refer
  to the HSS specification.

 **************************************************************************/

result = iqr_HSSSign(private_key, message, message_size, state, signature,
    signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Verifying Signatures

To verify a signature using the HSS public key:

...
result = iqr_HSSVerify(public_key, message, message_size, signature,
    signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Managing Keys

Be sure you’re not using the same private key on different systems unless you’ve used iqr_HSSDetachState(), and that you’re not re-using states.

To export HSS keys for storage or transmission:

...
size_t public_key_data_size = 0;
result = iqr_HSSGetPublicKeySize(params, &public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_data = calloc(1, public_key_data_size);
if (public_key_data == NULL) {
    // Handle error.
}

result = iqr_HSSExportPublicKey(public_key, public_key_data,
    public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

size_t private_key_data_size = 0;
result = iqr_HSSGetPrivateKeySize(params, &private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *private_key_data = calloc(1, private_key_data_size);
if (private_key_data == NULL) {
    // Handle error.
}

result = iqr_HSSExportPrivateKey(private_key, private_key_data,
    private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

size_t state_size = 0;
result = iqr_HSSGetStateSize(params, &state_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *state_data = calloc(1, state_size);
if (state_data == NULL) {
    // Handle error.
}

result = iqr_HSSExportState(state, state_data, state_size);
if (result != IQR_OK) {
    // Handle error.
}

To import HSS keys from buffers:

...
// Create an iqr_HSSParams object using the same parameters that were used
// to create the keys.
...

iqr_HSSPublicKey *public_key = NULL;
result = iqr_HSSImportPublicKey(params, public_key_data, public_key_data_size,
    &public_key);
if (result != IQR_OK) {
    // Handle error.
}

iqr_HSSPrivateKey *private_key = NULL;
result = iqr_HSSImportPrivateKey(params, private_key_data,
    private_key_data_size, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

iqr_HSSPrivateKeyState *state = NULL;
result = iqr_HSSImportState(params, state_data, state_data_size, &state);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for importing keys must match the `params` used to create the keys.

Public Key Format

HSS public keys are stored in the format defined in RFC 8554.

Signature Format

HSS signatures are stored in the format defined in RFC 8554.

SPHINCS+ Signature Scheme

The toolkit provides an implementation of the SPHINCS+ signature scheme as defined in the SPHINCS+ document.

SPHINCS+ is a stateless hash based signature scheme. Unlike HSS and XMSS, there’s no state that needs to be carefully maintained while signing messages.

The toolkit currently supports the following variants, using either SHA2-256 or SHAKE-256 as the internal hash:

IQR_SPHINCS_SHA2_256_128F — 128 bits of security, fast variant
IQR_SPHINCS_SHA2_256_128S — 128 bits of security, small variant
IQR_SPHINCS_SHAKE_256_128F — 128 bits of security, fast variant
IQR_SPHINCS_SHAKE_256_128S — 128 bits of security, small variant
IQR_SPHINCS_SHA2_256_192F — 192 bits of security, fast variant
IQR_SPHINCS_SHA2_256_192S — 192 bits of security, small variant
IQR_SPHINCS_SHAKE_256_192F — 192 bits of security, fast variant
IQR_SPHINCS_SHAKE_256_192S — 192 bits of security, small variant
IQR_SPHINCS_SHA2_256_256F — 256 bits of security, fast variant
IQR_SPHINCS_SHA2_256_256S — 256 bits of security, small variant
IQR_SPHINCS_SHAKE_256_256F — 256 bits of security, fast variant
IQR_SPHINCS_SHAKE_256_256S — 256 bits of security, small variant

The fast variants sign messages faster in exchange for larger signatures. The small variants sign slower but produce smaller signatures. See SPHINCS+ keys and signatures for sizes.

Signature verification speeds are unaffected by the variant.

Creating Keys

The toolkit lets you create SPHINCS+ keys after creating and initializing a random number generator.

To create a SPHINCS+ key pair:

#include "iqr_context.h"
#include "iqr_rng.h"
#include "iqr_sphincs.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// Create SPHINCS+ parameters using one of the variants from iqr_sphincs.h.
iqr_SPHINCSParams *params = NULL;
iqr_retval result = iqr_SPHINCSCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

// Create the key pair.
iqr_SPHINCSPublicKey *public_key = NULL;
iqr_SPHINCSPrivateKey *private_key = NULL;
result = iqr_SPHINCSCreateKeyPair(params, rng, &public_key, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

Signing

To sign a message using the SPHINCS+ private key:

...
size_t signature_size = 0;
result = iqr_SPHINCSGetSignatureSize(params, &signature_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *signature = calloc(1, signature_size);
if (signature == NULL) {
    // Handle error.
}

result = iqr_SPHINCSSign(private_key, rng, digest, digest_size,
    signature, signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Verifying Signatures

To verify a signature using the SPHINCS+ public key:

...
result = iqr_SPHINCSVerify(public_key, digest, digest_size, signature,
    signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Managing Keys

To export SPHINCS+ keys for storage or transmission:

...
size_t public_key_data_size = 0;
result = iqr_SPHINCSGetPublicKeySize(params, &public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_data = calloc(1, public_key_data_size);
if (public_key_data == NULL) {
    // Handle error.
}

result = iqr_SPHINCSExportPublicKey(public_key, public_key_data,
    public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

size_t private_key_data_size = 0;
result = iqr_SPHINCSGetPrivateKeySize(params, &private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *private_key_data = calloc(1, private_key_data_size);
if (private_key_data == NULL) {
    // Handle error.
}

result = iqr_SPHINCSExportPrivateKey(private_key, private_key_data,
    private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

To import SPHINCS+ keys from buffers:

...
// Create an iqr_SPHINCSParams object.
...

iqr_SPHINCSPublicKey *public_key = NULL;
result = iqr_SPHINCSImportPublicKey(params, public_key_data,
    public_key_data_size, &public_key);
if (result != IQR_OK) {
    // Handle error.
}

iqr_SPHINCSPrivateKey *private_key = NULL;
result = iqr_SPHINCSImportPrivateKey(params, private_key_data,
    private_key_data_size, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for importing keys must match the `params` used to create the keys.

Public Key Format

The data produced by iqr_SPHINCSExportPublicKey() follows the format specified in the SPHINCS+ document.

Signature Format

The signature produced by iqr_SPHINCSSign() follows the format specified in the SPHINCS+ document.

XMSS (eXtended Merkle Signature Scheme)

The toolkit provides an implementation of the eXtended Merkle Signature Scheme (XMSS) scheme as defined in the XMSS: eXtended Merkle Signature Scheme document.

XMSS is a one-time signature scheme that has several major differences from classical digital signature schemes:

An XMSS private key can only be used to sign a finite number of items.
You need to carefully maintain the state.

When you create an XMSS key pair, you specify the variant (iqr_XMSSVariant) parameter, which specifies the tree height. The height controls the number of one-time signatures available in the private key. The variant can be one of:

IQR_XMSS_2E10 — 2¹⁰ (1024) one-time signatures
IQR_XMSS_2E16 — 2¹⁶ (65,536) one-time signatures

It’s up to the user to manage domain parameters.

The state, specified when signing, tracks the one-time signature tree state. Re-using a one-time signature destroys the security of the scheme, so be careful to:

Not re-use a state when signing.
Safely and securely store your state before publishing the signature to protect against software crashes or power loss.

XMSS state is larger depending on the tree height; see the Technical Information section for sizes. XMSS signatures grow relative to the full tree height.

Creating Keys

The toolkit lets you create XMSS keys by specifying individual parameters.

To create an XMSS key pair:

#include "iqr_context.h"
#include "iqr_hash.h"
#include "iqr_rng.h"
#include "iqr_xmss.h"
...
// Create an iqr_Context, context.
// Register a SHA2-256 hash algorithm.
// Create and initialize a Random Number Generator, rng.
...
// Create XMSS parameters.
iqr_XMSSParams *params = NULL;
iqr_retval result = iqr_XMSSCreateParams(context, &IQR_XMSS_FULL_TREE_STRATEGY,
    variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

// Create the key pair.
iqr_XMSSPublicKey *public_key = NULL;
iqr_XMSSPrivateKey *private_key = NULL;
iqr_XMSSPrivateKeyState *state = NULL;
result = iqr_XMSSCreateKeyPair(params, rng, &public_key, &private_key,
    &state);
if (result != IQR_OK) {
    // Handle error.
}

Detaching State

Important

If this feature is used to move private data between two signing devices, it violates the conditions defined in NIST SP 800-208.

This is accomplished by detaching the key state. Using iqr_XMSSDetachState() will create a new state containing the number of signatures requested, and remove those signatures from the original state.

To detach num_sigs one-time signatures from an XMSS key’s state:

#include "iqr_xmss.h"
...
// This assumes you've already got an XMSS private key and state, and that
// you've registered a SHA2-256 hash implementation.
...
iqr_XMSSPrivateKeyState *new_state = NULL;
iqr_retval result = iqr_XMSSDetachState(private_key, original_state,
    num_sigs, &new_state)
if (result != IQR_OK) {
    // Handle error.
}

After calling iqr_XMSSDetachState(), num_sigs one-time signatures are removed from original_state. The new_state will have the removed signatures.

Signing

To sign a message using the XMSS private key:

...
size_t signature_size = 0;
result = iqr_XMSSGetSignatureSize(params, &signature_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *signature = calloc(1, signature_size);
if (signature == NULL) {
    // Handle error.
}

/*********************** CRITICALLY IMPORTANT STEP *************************

  You must detach a state to use for signing, and write the remaining state
  to non-volatile memory, before signing. Failure to do so could result in
  a security breach as it could lead to the re-use of a one-time signature.

  This step has been omitted for brevity.

  For more information about this property of the XMSS state, please refer
  to the XMSS specification.

 **************************************************************************/

result = iqr_XMSSSign(private_key, message, message_size, state, signature,
    signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Verifying Signatures

To verify a signature using the XMSS public key:

...
result = iqr_XMSSVerify(public_key, message, message_size, signature,
    signature_size);
if (result != IQR_OK) {
    // Handle error.
}

Managing Keys

Be sure you’re not using the same private key on different systems unless you’ve used iqr_XMSSDetachState(), and that you’re not re-using states.

To export XMSS keys for storage or transmission:

...
size_t public_key_data_size = 0;
result = iqr_XMSSGetPublicKeySize(params, &public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_data = calloc(1, public_key_data_size);
if (public_key_data == NULL) {
    // Handle error.
}

result = iqr_XMSSExportPublicKey(public_key, public_key_data,
    public_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

size_t private_key_data_size = 0;
result = iqr_XMSSGetPrivateKeySize(params, &private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *private_key_data = calloc(1, private_key_data_size);
if (private_key_data == NULL) {
    // Handle error.
}

result = iqr_XMSSExportPrivateKey(private_key, private_key_data,
    private_key_data_size);
if (result != IQR_OK) {
    // Handle error.
}

size_t state_size = 0;
result = iqr_XMSSGetStateSize(params, &state_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *state_data = calloc(1, state_size);
if (state_data == NULL) {
    // Handle error.
}

result = iqr_XMSSExportState(state, state_data, state_size);
if (result != IQR_OK) {
    // Handle error.
}

To import XMSS keys from buffers:

...
// Create an iqr_XMSSParams object using the same parameters that were used
// to create the keys.
...

iqr_XMSSPublicKey *public_key = NULL;
result = iqr_XMSSImportPublicKey(params, public_key_data, public_key_data_size,
    &public_key);
if (result != IQR_OK) {
    // Handle error.
}

iqr_XMSSPrivateKey *private_key = NULL;
result = iqr_XMSSImportPrivateKey(params, private_key_data,
    private_key_data_size, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

iqr_XMSSPrivateKeyState *state = NULL;
result = iqr_XMSSImportState(params, state_data, state_data_size, &state);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for importing keys must match the `params` used to create the keys.

Public Key Format

The data produced by iqr_XMSSExportPublicKey() follows the format documented in section 4.1.7 of the XMSS RFC.

Supported oid values for the toolkit match these values from section 8 of the IETF document:

XMSS-SHA2_10_256
XMSS-SHA2_16_256

Signature Format

The signature produced by iqr_XMSSSign() follows the format documented in section 4.1.8 of the XMSS RFC.

Key Agreement

The toolkit provides these key agreement schemes:

FrodoDH — a lattice-based key agreement scheme
Samwise — a lattice-based key agreement scheme
SIDH (Supersingular-Isogeny Diffie-Hellman) — an isogeny-based key agreement scheme

FrodoDH and Samwise differ from standard Diffie-Hellman. In these schemes, whoever starts the key exchange protocol is the Initiator, and the other is the Responder. The Initiator and the Responder then exchange keys in the following modified manner:

The Initiator and the Responder agree on a set of public parameters.
The Initiator creates a public/private key pair.
The Initiator sends their public key to the Responder.
The Responder creates a private key, then uses it and the Initiator’s public key to create their own public key.
The Responder sends their public key to the Initiator.
The Initiator uses their private key and the Responder’s public key to calculate the shared secret.
The Responder uses their private key to calculate the shared secret.

The last two steps can occur simultaneously.

SIDH also differs from standard Diffie-Hellman. In SIDH, one end of the connection must be Alice, the other Bob. It doesn’t matter which is which, as long as they’re not the same.

Alice and Bob agree on a set of public parameters.
Alice creates a private key.
Alice uses her private key to derive a public key.
Alice sends the public key to Bob.
Bob creates a private key.
Bob uses his private key to derive a public key.
Bob sends the public key to Alice.
Alice uses her private key and Bob’s public key to derive the shared secret.
Bob uses his private key and Alice’s public key to calculate the shared secret.

Steps 2-4 for Alice and 5-7 for Bob can be done simultaneously, since the calculations are done independent of one another.

Note	The shared secrets generated by these key agreement schemes are ephemeral. You cannot re-use keys to generate additional shared secrets.

Key Agreement Usage

Decide who gets to be Alice (or the Initiator), and who gets to be Bob (or the Responder).
Alice creates a new private key and uses it to derive a public key.
- Alice and Bob must create a new private key and public key every time.
Alice sends her public key to Bob.
Bob creates a new private key and uses it, plus Alice’s public key, to derive his own public key.
Bob uses his private key to derive the shared secret.
Bob sends his public key to Alice.
Alice uses her private key and Bob’s public key to derive the shared secret.
Feed the shared secret to a Key Derivation Function to get a shared key.
Use the shared key for symmetric encryption (for example).

Note	After the agreement, be sure to destroy the keys. Depending on the algorithm, it might not be safe to re-use them for additional key agreements.

Perfect Forward Secrecy

To preserve perfect forward secrecy, use a new private key for each transaction. Treat them as ephemeral.

To help enforce this, the private key Export() and Import() functions will be removed in a future version of the toolkit.

FrodoDH

The toolkit provides the following variants of the FrodoDH lattice-based key agreement scheme:

IQR_FRODODH_976_AES
IQR_FRODODH_976_SHAKE

Depending on the protocol you’re implementing, either side of the agreement could be acting as Initiator or Responder.

It’s up to the user to manage domain parameters; the parameter data is not exposed in stored keys or shared secrets.

FrodoDH’s variants choose the algorithm used internally to generate a pseudo-random function (AES or SHAKE). Both variants provide the same level of security.

Note	The Initiator and Responder must use the same parameters.

Initiator

To create a key pair:

#include "iqr_context.h"
#include "iqr_frododh.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be &IQR_FRODODH_976_AES or &IQR_FRODODH_976_SHAKE
// depending on which implementation the Responder is using.
iqr_FrodoDHParams *params = NULL;
iqr_retval result = iqr_FrodoDHCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_FrodoDHInitiatorPrivateKey *private_key = NULL;
result = iqr_FrodoDHCreateInitiatorPrivateKey(params, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_FrodoDHGetInitiatorPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_FrodoDHGetInitiatorPublicKey(private_key, rng, public_key,
        public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key is now sent to the Responder.
...

To calculate the shared secret using the Responder’s public key:

...
size_t responder_public_key_size = 0;
result = iqr_FrodoDHGetResponderPublicKeySize(params, &responder_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *responder_public_key = calloc(1, responder_public_key_size);
if (responder_public_key == NULL) {
    // Handle error.
}

// Receive the Responder's public key.
...

size_t shared_secret_size = 0;
result = iqr_FrodoDHGetSecretSize(prams, &shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_secret = calloc(1, shared_secret_size);
if (shared_secret == NULL) {
    // Handle error.
}

result = iqr_FrodoDHGetInitiatorSecret(private_key, responder_public_key,
    responder_public_key_size, shared_secret, shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

Responder

To create a key pair using the Initiator’s public key:

#include "iqr_context.h"
#include "iqr_frododh.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be &IQR_FRODODH_976_AES or &IQR_FRODODH_976_SHAKE
// depending on which implementation the Initiator is using.
iqr_FrodoDHParams *params = NULL;
iqr_retval result = iqr_FrodoDHCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

size_t initiator_public_key_size = 0;
result = iqr_FrodoDHGetInitiatorPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *initiator_public_key = calloc(1, initiator_public_key_size);
if (inititator_public_key == NULL) {
    // Handle error.
}

// Receive the Initiator's public key.
...

iqr_FrodoDHResponderPrivateKey *private_key = NULL;
result = iqr_FrodoDHCreateResponderPrivateKey(params, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t responder_public_key_size = 0;
result = iqr_FrodoDHGetResponderPublicKeySize(params, &responder_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

result = iqr_FrodoDHGetResponderPublicKey(private_key, rng,
        initiator_public_key, initiator_public_key_size, responder_public_key,
        responder_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// Send the Responder's public key to the Initiator.
...

To calculate the shared secret using the Responder’s private key:

...
size_t shared_secret_size = 0;
result = iqr_FrodoDHGetSecretSize(params, &shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_secret = calloc(1, shared_secret_size);
if (shared_secret == NULL) {
    // Handle error.
}

result = iqr_FrodoDHGetResponderSecret(private_key, shared_secret,
        shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}
...

Samwise

The toolkit provides the following variants of the Samwise lattice-based key agreement scheme:

IQR_SAMWISE_976_AES
IQR_SAMWISE_976_CHACHA20

Depending on the protocol you’re implementing, either side of the agreement could be acting as Initiator or Responder.

It’s up to the user to manage domain parameters; the parameter data is not exposed in stored keys or shared secrets.

Samwise’s variants choose the algorithm used internally to generate a pseudo-random function (AES or ChaCha20). Both variants provide the same level of security.

Samwise is an optimized variant of FrodoDH designed by ISARA.

Note	The Initiator and Responder must use the same parameters.

Initiator

To create a key pair:

#include "iqr_context.h"
#include "iqr_samwise.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be &IQR_SAMWISE_976_AES or &IQR_SAMWISE_976_CHACHA20
// depending on which implementation the Responder is using.
iqr_SamwiseParams *params = NULL;
iqr_retval result = iqr_SamwiseCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_SamwiseInitiatorPrivateKey *private_key = NULL;
result = iqr_SamwiseCreateInitiatorPrivateKey(params, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_SamwiseGetInitiatorPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_SamwiseGetInitiatorPublicKey(private_key, rng, public_key,
        public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key is now sent to the Responder.
...

To calculate the shared secret using the Responder’s public key:

...
size_t responder_public_key_size = 0;
result = iqr_SamwiseGetResponderPublicKeySize(params, &responder_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *responder_public_key = calloc(1, responder_public_key_size);
if (responder_public_key == NULL) {
    // Handle error.
}

// Receive the Responder's public key.
...

size_t shared_secret_size = 0;
result = iqr_SamwiseGetSecretSize(prams, &shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_secret = calloc(1, shared_secret_size);
if (shared_secret == NULL) {
    // Handle error.
}

result = iqr_SamwiseGetInitiatorSecret(private_key, responder_public_key,
    responder_public_key_size, shared_secret, shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

Responder

To create a key pair using the Initiator’s public key:

#include "iqr_context.h"
#include "iqr_samwise.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be &IQR_SAMWISE_976_AES or &IQR_SAMWISE_976_CHACHA20
// depending on which implementation the Initiator is using.
iqr_SamwiseParams *params = NULL;
iqr_retval result = iqr_SamwiseCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

size_t initiator_public_key_size = 0;
result = iqr_SamwiseGetInitiatorPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *initiator_public_key = calloc(1, initiator_public_key_size);
if (inititator_public_key == NULL) {
    // Handle error.
}

// Receive the Initiator's public key.
...

iqr_SamwiseResponderPrivateKey *private_key = NULL;
result = iqr_SamwiseCreateResponderPrivateKey(params, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t responder_public_key_size = 0;
result = iqr_SamwiseGetResponderPublicKeySize(params, &responder_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

result = iqr_SamwiseGetResponderPublicKey(private_key, rng,
        initiator_public_key, initiator_public_key_size, responder_public_key,
        responder_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// Send the Responder's public key to the Initiator.
...

To calculate the shared secret using the Responder’s private key:

...
size_t shared_secret_size = 0;
result = iqr_SamwiseGetSecretSize(params, &shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_secret = calloc(1, shared_secret_size);
if (shared_secret == NULL) {
    // Handle error.
}

result = iqr_SamwiseGetResponderSecret(private_key, shared_secret,
        shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}
...

SIDH

The toolkit provides the following variants of the SIDH isogeny-based key agreement scheme:

IQR_SIDH_P434 - Parameters using the p434 curve.
IQR_SIDH_P503 - Parameters using the p503 curve.
IQR_SIDH_P610 - Parameters using the p610 curve.
IQR_SIDH_P751 - Parameters using the p751 curve.

Depending on the protocol you’re implementing, either side of the agreement could be acting as Alice or Bob. It doesn’t matter which is which, as long as they’re not the same. Unlike the lattice-based key agreement schemes, Alice and Bob can generate their keys simultaneously, without input from the other side of the transaction.

It’s up to the user to manage domain parameters; the parameter data is not exposed in stored keys or shared secrets.

The p434 curve provides less security in exchange for better performance. The p751 curve provides the greatest security with the biggest performance cost. The best security/performance trade-off is with the p503 curve.

Note	Alice and Bob must use the same parameters.

Alice

To create a key pair for Alice:

#include "iqr_context.h"
#include "iqr_rng.h"
#include "iqr_sidh.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// Create SIDH parameters using one of the variants from iqr_sidh.h.
iqr_SIDHParams *params = NULL;
iqr_retval result = iqr_SIDHCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_SIDHAlicePrivateKey *alice_private_key = NULL;
result = iqr_SIDECreateAlicePrivateKey(params, rng, &alice_private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t alice_public_key_size = 0;
result = iqr_SIDHGetPublicKeySize(params, &alice_public_key_size);
uint8_t *alice_public_key = calloc(1, alice_public_key_size);
if (alice_public_key == NULL) {
    // Handle error.
}

result = iqr_SIDHGetAlicePublicKey(alice_private_key,
    alice_public_key, alice_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The alice_public_key is now sent to Bob.
...

To calculate the shared secret using Bob’s public key:

...
size_t bob_public_key_size = 0;
result = iqr_SIDHGetPublicKeySize(params, &bob_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *bob_public_key = calloc(1, bob_public_key_size);
if (bob_public_key == NULL) {
    // Handle error.
}

// Receive Bob's public key.
...

size_t shared_secret_size = 0;
result = iqr_SIDHGetSecretSize(params, &shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_secret = calloc(1, shared_secret_size);
if (shared_secret == NULL) {
    // Handle error.
}

result = iqr_SIDHGetAliceSecret(alice_private_key, bob_public_key,
    bob_public_key_size, shared_secret, shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

Bob

To create a key pair for Bob:

#include "iqr_context.h"
#include "iqr_rng.h"
#include "iqr_sidh.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
iqr_SIDHParams *params = NULL;
iqr_retval result = iqr_SIDHCreateParams(context, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_SIDHBobPrivateKey *bob_private_key = NULL;
result = iqr_SIDECreateBobPrivateKey(params, rng, &bob_private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t bob_public_key_size = 0;
result = iqr_SIDHGetPublicKeySize(params, &bob_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *bob_public_key = calloc(1, bob_public_key_size);
if (bob_public_key == NULL) {
    // Handle error.
}

result = iqr_SIDHGetBobPublicKey(bob_private_key,
    bob_public_key, bob_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The bob_public_key is now sent to Alice.
...

To calculate the shared secret using Alice’s public key:

...
size_t alice_public_key_size = 0;
result = iqr_SIDHGetPublicKeySize(params, &alice_public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *alice_public_key = calloc(1, alice_public_key_size);
if (alice_public_key == NULL) {
    // Handle error.
}

// Receive Alice's public key.
...

size_t shared_secret_size = 0;
result = iqr_SIDHGetSecretSize(params, &shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_secret = calloc(1, shared_secret_size);
if (shared_secret == NULL) {
    // Handle error.
}

result = iqr_SIDHGetBobSecret(bob_private_key, alice_public_key,
    alice_public_key_size, shared_secret, shared_secret_size);
if (result != IQR_OK) {
    // Handle error.
}

Key Encapsulation

The toolkit provides these key encapsulation mechanisms (KEMs):

Classic McEliece — a code-based KEM
FrodoKEM — a lattice-based KEM
Kyber — a lattice-based KEM
NTRUPrime — a lattice-based KEM
Saber — a lattice-based KEM
SIKE — (Supersingular-Isogeny Key Encapsulation) — an isogeny-based KEM

These schemes let you create and cryptographically encapsulate a symmetric key for safe transmission. They work similar to public key encryption schemes.

To show how Alice and Bob would use a KEM:

Bob generates a KEM key pair.
Bob sends the KEM public key to Alice.
Alice uses Bob’s public key to generate a shared secret and a ciphertext.
Alice sends the ciphertext to Bob.
Bob uses the KEM’s private key to decapsulate the ciphertext, giving him the original shared secret.
Alice and Bob independently derive an encryption key from the shared secret.

Note	If your protocol requires perfect forward secrecy, you must treat the KEM private keys as ephemeral keys; don’t re-use them.

Key Encapsulation Mechanism Usage

Alice generates a key pair.
- Alice makes her public key available to anyone.
- Alice keeps her private key secret.
Bob uses Alice’s public key to generate a shared secret and encapsulate it as a ciphertext.
- Bob sends the ciphertext back to Alice.
Alice uses her private key and the ciphertext to decapsulate the shared secret.
- When they’re done, both Alice and Bob have the same shared secret.
Use the shared secret for symmetric encryption (for example).

Depending on your protocol, you might also feed the shared secret to a Key Derivation Function to get a shared encryption key.

Classic McEliece

The toolkit provides the following variants of the Classic McEliece KEM:

IQR_CLASSICMCELIECE_6 — The "6960119" variant from the specification
IQR_CLASSICMCELIECE_8 — The "8192128" variant from the specification

It’s up to the user to manage domain parameters (the variant); the parameter data is not exposed in stored keys or ciphertexts.

Both provide the same general level of security. IQR_CLASSICMCELIECE_8 produces larger keys and ciphertext and requires more CPU resources in exchange for stronger security.

Creating Keys

To create a key pair:

#include "iqr_classicmceliece.h"
#include "iqr_context.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be IQR_CLASSICMCELIECE_6 or IQR_CLASSICMCELIECE_8.
iqr_ClassicMcElieceParams *params = NULL;
iqr_retval result = iqr_ClassicMcElieceCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_ClassicMcEliecePrivateKey *private_key = NULL;
iqr_ClassicMcEliecePublicKey *public_key = NULL;
result = iqr_ClassicMcElieceCreateKeyPair(params, rng, &public_key,
    &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_ClassicMcElieceGetPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_buffer = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_ClassicMcElieceExportPublicKey(public_key, public_key_buffer,
    public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key can now be sent to other users.
...

Encapsulating the Secret

To encapsulate a shared secret using the public key:

size_t shared_key_size = 0;
result = iqr_ClassicMcElieceGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

size_t ciphertext_size = 0;
result = iqr_ClassicMcElieceGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *ciphertext = calloc(1, ciphertext_size);
if (ciphertext == NULL) {
    // Handle error.
}

result = iqr_ClassicMcElieceEncapsulate(public_key, rng, ciphertext,
    ciphertext_size, shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The ciphertext can now be sent to the recipient.
...

Decapsulating the Secret

To decapsulate the shared secret using the private key:

// Get the ciphertext from the sender.
...

size_t ciphertext_size = 0;
result = iqr_ClassicMcElieceGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

if (ciphertext_size != received_ciphertext_size) {
    // Handle error.
}

size_t shared_key_size = 0;
result = iqr_ClassicMcElieceGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

result = iqr_ClassicMcElieceDecapsulate(private_key, ciphertext,
    ciphertext_size, shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for decapsulating the ciphertext must match the `params` used to create the keys.

FrodoKEM

The toolkit provides the following variants of FrodoKEM:

IQR_FRODOKEM_640_AES — 104 bits of quantum security with AES for internal permutations
IQR_FRODOKEM_640_SHAKE — 104 bits of quantum security with SHAKE for internal permutations
IQR_FRODOKEM_976_AES — 150 bits of quantum security with AES for internal permutations
IQR_FRODOKEM_976_SHAKE — 150 bits of quantum security with SHAKE for internal permutations
IQR_FRODOKEM_1344_AES — 197 bits of quantum security with AES for internal permutations
IQR_FRODOKEM_1344_SHAKE — 197 bits of quantum security with SHAKE for internal permutations

It’s up to the user to manage domain parameters (the variant); the parameter data is not exposed in stored keys or ciphertexts.

FrodoKEM’s variants choose the algorithm used internally to generate a pseudo-random function (AES or SHAKE). Both sets of variants provide the same level of security.

Creating Keys

To create a key pair:

#include "iqr_context.h"
#include "iqr_frodokem.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be any of the IQR_FRODOKEM_* variants in iqr_frodokem.h.
iqr_FrodoKEMParams *params = NULL;
iqr_retval result = iqr_FrodoKEMCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_FrodoKEMPrivateKey *private_key = NULL;
iqr_FrodoKEMPublicKey *public_key = NULL;
result = iqr_FrodoKEMCreateKeyPair(params, rng, &public_key, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_FrodoKEMGetPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_buffer = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_FrodoKEMExportPublicKey(public_key, public_key_buffer, public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key can now be sent to other users.
...

Encapsulating the Secret

To encapsulate a shared secret using the public key:

size_t shared_key_size = 0;
result = iqr_FrodoKEMGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

size_t ciphertext_size = 0;
result = iqr_FrodoKEMGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *ciphertext = calloc(1, ciphertext_size);
if (ciphertext == NULL) {
    // Handle error.
}

result = iqr_FrodoKEMEncapsulate(public_key, rng, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The ciphertext can now be sent to the recipient.
...

Decapsulating the Secret

To decapsulate the shared secret using the private key:

// Get the ciphertext from the sender.
...

size_t ciphertext_size = 0;
result = iqr_FrodoKEMGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

if (ciphertext_size != received_ciphertext_size) {
    // Handle error.
}

size_t shared_key_size = 0;
result = iqr_FrodoKEMGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

result = iqr_FrodoKEMDecapsulate(private_key, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for decapsulating the ciphertext must match the `params` used to create the keys.

Kyber

The toolkit provides the following variants of the Kyber KEM:

IQR_KYBER_512 — 100 bits of quantum security
IQR_KYBER_768 — 160 bits of quantum security
IQR_KYBER_1024 — 224 bits of quantum security

It’s up to the user to manage domain parameters (the variant); the parameter data is not exposed in stored keys or ciphertexts.

Creating Keys

To create a key pair:

#include "iqr_context.h"
#include "iqr_kyber.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be IQR_KYBER_512, IQR_KYBER_768, or IQR_KYBER_1024.
iqr_KyberParams *params = NULL;
iqr_retval result = iqr_KyberCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_KyberPrivateKey *private_key = NULL;
iqr_KyberPublicKey *public_key = NULL;
result = iqr_KyberCreateKeyPair(params, rng, &public_key, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_KyberGetPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_buffer = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_KyberExportPublicKey(public_key, public_key_buffer, public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key can now be sent to other users.
...

Encapsulating the Secret

To encapsulate a shared secret using the public key:

size_t shared_key_size = 0;
result = iqr_KyberGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

size_t ciphertext_size = 0;
result = iqr_KyberGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *ciphertext = calloc(1, ciphertext_size);
if (ciphertext == NULL) {
    // Handle error.
}

result = iqr_KyberEncapsulate(public_key, rng, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The ciphertext can now be sent to the recipient.
...

Decapsulating the Secret

To decapsulate the shared secret using the private key:

// Get the ciphertext from the sender.
...

size_t ciphertext_size = 0;
result = iqr_KyberGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

if (ciphertext_size != received_ciphertext_size) {
    // Handle error.
}

size_t shared_key_size = 0;
result = iqr_KyberGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

result = iqr_KyberDecapsulate(private_key, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for decapsulating the ciphertext must match the `params` used to create the keys.

NTRUPrime

The toolkit provides the following variants of the NTRUPrime KEM:

IQR_SNTRUP_653 — NIST category 1 variant
IQR_SNTRUP_761 — NIST category 2 variant
IQR_SNTRUP_857 — NIST category 2-3 variant
IQR_SNTRUP_953 — NIST category 3-4 variant
IQR_SNTRUP_1013 — NIST category 4 variant
IQR_SNTRUP_1277 — NIST category 5 variant

It’s up to the user to manage domain parameters (the variant); the parameter data is not exposed in stored keys or ciphertexts.

Creating Keys

To create a key pair:

#include "iqr_context.h"
#include "iqr_hash.h"
#include "iqr_ntruprime.h"
#include "iqr_rng.h"
...
// Create an iqr_Context, context.
// Register a SHA2-512 hash implementation.
// Create and initialize a Random Number Generator, rng.
...
iqr_NTRUPrimeParams *params = NULL;
iqr_retval result = iqr_NTRUPrimeCreateParams(context, &IQR_SNTRUP_761, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_NTRUPrimePrivateKey *private_key = NULL;
iqr_NTRUPrimePublicKey *public_key = NULL;
result = iqr_NTRUPrimeCreateKeyPair(params, rng, &public_key, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_NTRUPrimeGetPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_buffer = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_NTRUPrimeExportPublicKey(public_key, public_key_buffer,
    public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key can now be sent to other users.
...

Encapsulating the Secret

To encapsulate a shared secret using the public key:

size_t shared_key_size = 0;
result = iqr_NTRUPrimeGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

size_t ciphertext_size = 0;
result = iqr_NTRUPrimeGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *ciphertext = calloc(1, ciphertext_size);
if (ciphertext == NULL) {
    // Handle error.
}

result = iqr_NTRUPrimeEncapsulate(public_key, rng, ciphertext,
    ciphertext_size, shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The ciphertext can now be sent to the recipient.
...

Decapsulating the Secret

To decapsulate the shared secret using the private key:

// Get the ciphertext from the sender.
...

size_t ciphertext_size = 0;
result = iqr_NTRUPrimeGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

if (received_ciphertext_size != ciphertext_size) {
    // Handle error.
}

size_t shared_key_size = 0;
result = iqr_NTRUPrimeGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

result = iqr_NTRUPrimeDecapsulate(private_key, ciphertext,
    ciphertext_size, shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for decapsulating the ciphertext must match the `params` used to create the keys.

Saber

The toolkit provides the following variants of the Saber lattice-based KEM:

IQR_LIGHT_SABER — Parameters for the LightSaber variant
IQR_SABER — Parameters for the Saber variant
IQR_FIRE_SABER — Parameters for the FireSaber variant

It’s up to the user to manage domain parameters (the variant); the parameter data is not exposed in stored keys or ciphertexts.

The LightSaber variant trades lower security for faster computing, while the FireSaber variant gives you higher security at the cost of more computation.

Creating Keys

To create a key pair:

#include "iqr_context.h"
#include "iqr_rng.h"
#include "iqr_saber.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be IQR_LIGHT_SABER, IQR_SABER, or IQR_FIRE_SABER.
iqr_SaberParams *params = NULL;
iqr_retval result = iqr_SaberCreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_SaberPrivateKey *private_key = NULL;
iqr_SaberPublicKey *public_key = NULL;
result = iqr_SaberCreateKeyPair(params, rng, &public_key, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_SaberGetPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_buffer = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_SaberExportPublicKey(public_key, public_key_buffer, public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key can now be sent to other users.
...

Encapsulating the Key

To encapsulate a shared key using the public key:

size_t shared_key_size = 0;
result = iqr_SaberGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

size_t ciphertext_size = 0;
result = iqr_SaberGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *ciphertext = calloc(1, ciphertext_size);
if (ciphertext == NULL) {
    // Handle error.
}

result = iqr_SaberEncapsulate(public_key, rng, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The ciphertext can now be sent to the recipient.
...

Decapsulating the Key

To decapsulate the shared key using the private key:

// Get the ciphertext from the sender.
...

size_t ciphertext_size = 0;
result = iqr_SaberGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

if (ciphertext_size != received_ciphertext_size) {
    // Handle error.
}

size_t shared_key_size = 0;
result = iqr_SaberGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

result = iqr_SaberDecapsulate(private_key, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for decapsulating the ciphertext must match the `params` used to create the keys.

SIKE

The toolkit provides the following variants of the SIKE isogeny-based KEM:

IQR_SIKE_P434 — Parameters using the NIST p434 curve
IQR_SIKE_P503 — Parameters using the NIST p503 curve
IQR_SIKE_P610 — Parameters using the NIST p610 curve
IQR_SIKE_P751 — Parameters using the NIST p751 curve

It’s up to the user to manage domain parameters (the variant); the parameter data is not exposed in stored keys or ciphertexts.

Creating Keys

To create a key pair:

#include "iqr_context.h"
#include "iqr_rng.h"
#include "iqr_sike.h"
...
// Create an iqr_Context, context.
// Create and initialize a Random Number Generator, rng.
...
// The variant can be IQR_SIKE_P434, IQR_SIKE_P503, IQR_SIKE_P610,
// or IQR_SIKE_P751.
iqr_SIKEParams *params = NULL;
iqr_retval result = iqr_SIKECreateParams(context, variant, &params);
if (result != IQR_OK) {
    // Handle error.
}

iqr_SIKEPrivateKey *private_key = NULL;
iqr_SIKEPublicKey *public_key = NULL;
result = iqr_SIKECreateKeyPair(params, rng, &public_key, &private_key);
if (result != IQR_OK) {
    // Handle error.
}

size_t public_key_size = 0;
result = iqr_SIKEGetPublicKeySize(params, &public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *public_key_buffer = calloc(1, public_key_size);
if (public_key == NULL) {
    // Handle error.
}

result = iqr_SIKEExportPublicKey(public_key, public_key_buffer, public_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The public_key can now be sent to other users.
...

Encapsulating the Key

To encapsulate a shared key using the public key:

size_t shared_key_size = 0;
result = iqr_SIKEGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

size_t ciphertext_size = 0;
result = iqr_SIKEGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *ciphertext = calloc(1, ciphertext_size);
if (ciphertext == NULL) {
    // Handle error.
}

result = iqr_SIKEEncapsulate(public_key, rng, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

// The ciphertext can now be sent to the recipient.
...

Decapsulating the Key

To decapsulate the shared key using the private key:

// Get the ciphertext from the sender.
...

size_t ciphertext_size = 0;
result = iqr_SIKEGetCiphertextSize(params, &ciphertext_size);
if (result != IQR_OK) {
    // Handle error.
}

if (ciphertext_size != received_ciphertext_size) {
    // Handle error.
}

size_t shared_key_size = 0;
result = iqr_SIKEGetSharedKeySize(params, &shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

uint8_t *shared_key = calloc(1, shared_key_size);
if (shared_key == NULL) {
    // Handle error.
}

result = iqr_SIKEDecapsulate(private_key, ciphertext, ciphertext_size,
    shared_key, shared_key_size);
if (result != IQR_OK) {
    // Handle error.
}

Note	The `params` used for decapsulating the ciphertext must match the `params` used to create the keys.

Technical Information

This section provides further information about some technical aspects of the toolkit.

Estimated Security

Section 4 of the NIST Call for Proposals (CFP) for Post-Quantum Cryptography describes these requirements for security:

KEMs that satisfy IND-CCA2 ("Indistinguishable under adaptive chosen ciphertext attack") security can be safely used with static keys (up to 2⁶⁴ times).
KEMs that satisfy IND-CPA ("Indistinguishable under chosen plaintext attack") security can only be used with entirely ephemeral keys.
Digital signatures must satisfy EUF-CMA ("Existential Unforgeability under Chosen Message Attack") security (for up to 2⁶⁴ messages).

The CFP defines five security strengths:

I - Equivalent to finding the key for a block cipher with a 128-bit key (AES128).
II - Equivalent to finding a collision in a 256-bit hash function (SHA2-256, SHA3-256).
III - Equivalent to finding the key for a block cipher with a 192-bit key (AES192).
IV - Equivalent to finding a collision in a 384-bit hash function (SHA2-384, SHA3-384).
V - Equivalent to finding the key for a block cipher with a 256-bit key (AES256).

You can find more detailed information about these ratings in the CFP document.

Key Agreement Scheme Security

The NIST competition only applies to KEMs and digital signature schemes. The toolkit’s key agreement schemes are generally based on the KEM versions found in the competition.

Table 3. Key Agreement Security Strength
Algorithm	Variant	Strength
FrodoDH	`IQR_FRODODH_976_AES`	III
	`IQR_FRODODH_976_SHAKE`	III
Samwise	`IQR_SAMWISE_976_AES`	III
	`IQR_SAMWISE_976_CHACHA20`	III
SIDH	`IQR_SIDH_P434`	I
	`IQR_SIDH_P503`	II
	`IQR_SIDH_P610`	III
	`IQR_SIDH_P751`	V

Note	In general, the key agreement schemes should not be used with static keys, only ephemerally.

KEM Security

The toolkit’s KEMs support various security strengths depending on their variants:

Table 4. KEM Security Strength
Algorithm	Variant	Strength
Classic McEliece	`IQR_CLASSICMCELIECE_6`	V
	`IQR_CLASSICMCELIECE_8`	V
Frodo KEM	`IQR_FRODOKEM_640_AES`	I
	`IQR_FRODOKEM_640_SHAKE`	I
	`IQR_FRODOKEM_976_AES`	III
	`IQR_FRODOKEM_976_SHAKE`	III
	`IQR_FRODOKEM_1344_AES`	V
	`IQR_FRODOKEM_1344_SHAKE`	V
Kyber	`IQR_KYBER_512`	I
	`IQR_KYBER_768`	III
	`IQR_KYBER_1024`	V
NTRUPrime	`IQR_SNTRUP_653`	I
	`IQR_SNTRUP_761`	II
	`IQR_SNTRUP_857`	II - III
	`IQR_SNTRUP_953`	III - IV
	`IQR_SNTRUP_1013`	IV
	`IQR_SNTRUP_1277`	V
Saber	`IQR_LIGHT_SABER`	I
	`IQR_SABER`	III
	`IQR_FIRE_SABER`	V
SIKE	`IQR_SIKE_P434`	I
	`IQR_SIKE_P503`	II
	`IQR_SIKE_P610`	III
	`IQR_SIKE_P751`	V

Note	KEMs all provide IND-CCA2 security, and are safe to use with static keys.

Signature Scheme Security

The stateful hash-based signature schemes (HSS and XMSS) have security equivalent to their underlying hash function, SHA2-256. This is equivalent to NIST’s II security category.

The toolkit’s NIST signature schemes support various security strengths depending on their variants:

Table 5. Signature Scheme Security Strength
Algorithm	Variant	Strength
Dilithium	`IQR_DILITHIUM_2`	II
	`IQR_DILITHIUM_3`	III
	`IQR_DILITHIUM_5`	V
SPHINCS+	`IQR_SPHINCS_SHA2_256_128F`	I
	`IQR_SPHINCS_SHA2_256_128S`	I
	`IQR_SPHINCS_SHAKE_256_128F`	I
	`IQR_SPHINCS_SHAKE_256_128S`	I
	`IQR_SPHINCS_SHA2_256_192F`	III
	`IQR_SPHINCS_SHA2_256_192S`	III
	`IQR_SPHINCS_SHAKE_256_192F`	III
	`IQR_SPHINCS_SHAKE_256_192S`	III
	`IQR_SPHINCS_SHA2_256_256F`	V
	`IQR_SPHINCS_SHA2_256_256S`	V
	`IQR_SPHINCS_SHAKE_256_256F`	V
	`IQR_SPHINCS_SHAKE_256_256S`	V

Key and Signature Sizes

Digital signature schemes:

Dilithium keys and signatures
HSS keys and signatures
SPHINCS+ keys and signatures
XMSS keys and signatures

Key agreement schemes:

FrodoDH keys and secrets
Samwise keys and secrets
SIDH keys and secrets

Key encapsulation mechanisms:

Classic McEliece keys, ciphertexts, and secrets
FrodoKEM keys, ciphertexts, and secrets
Kyber keys, ciphertexts, and secrets
NTRUPrime keys, ciphertexts, and secrets
Saber keys, ciphertexts, and secrets
SIKE keys, ciphertexts, and secrets

Note	These sizes are provided as a reference only; use the various `iqr_GetSize()` functions to find the correct sizes at run-time.

Dilithium Keys and Signatures

Dilithium digital signature scheme key and signature sizes:

Table 6. Dilithium Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Signature (bytes)
`IQR_DILITHIUM_2`	1,312	13,672	2,420
`IQR_DILITHIUM_3`	1,952	19,432	3,293
`IQR_DILITHIUM_5`	2,592	26,216	4,595

HSS Keys and Signatures

HSS digital signature scheme key and signature sizes:

Table 7. HSS Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Memory Constrained State (bytes)	Full Tree State (bytes)	Signature (bytes)
`IQR_HSS_2E15_FAST`	60	120	7,568	2,101,988	4,784
`IQR_HSS_2E15_SMALL`	60	120	4,400	2,098,820	1,616

The tree strategy has no effect on HSS key or signature sizes.

SPHINCS+ Keys and Signatures

Table 8. SPHINCS+ Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Signature (bytes)
`IQR_SPHINCS_SHA2_256_128F`	32	72	17,088
`IQR_SPHINCS_SHA2_256_128S`	32	72	7,856
`IQR_SPHINCS_SHAKE_256_128F`	32	72	17,088
`IQR_SPHINCS_SHAKE_256_128S`	32	72	7,856
`IQR_SPHINCS_SHA2_256_192F`	48	104	35,664
`IQR_SPHINCS_SHA2_256_192S`	48	104	16,224
`IQR_SPHINCS_SHAKE_256_192F`	48	104	35,664
`IQR_SPHINCS_SHAKE_256_192S`	48	104	16,224
`IQR_SPHINCS_SHA2_256_256F`	64	136	49,856
`IQR_SPHINCS_SHA2_256_256S`	64	136	29,792
`IQR_SPHINCS_SHAKE_256_256F`	64	136	49,856
`IQR_SPHINCS_SHAKE_256_256S`	64	136	29,792

XMSS Keys and Signatures

XMSS digital signature scheme key and signature sizes:

Table 9. XMSS Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Memory Constrained State (bytes)	Full Tree State (bytes)	Signature (bytes)
`IQR_XMSS_2E10`	68	104	1,276	65,568	2,500
`IQR_XMSS_2E16`	68	104	3,152	4,194,336	2,692

The tree strategy has no effect on XMSS key or signature sizes.

FrodoDH Keys and Secrets

FrodoDH key and shared secret sizes:

Table 10. FrodoDH Sizes
Variant	Initiator Public Key (bytes)	Responder Public Key (bytes)	Shared Secret (bytes)
`IQR_FRODODH_976_AES`	15,632	15,624	24
`IQR_FRODODH_976_SHAKE`	15,632	15,624	24

Samwise Keys and Secrets

Samwise key and shared secret sizes:

Table 11. Samwise Sizes
Variant	Initiator Public Key (bytes)	Responder Public Key (bytes)	Shared Secret (bytes)
`IQR_SAMWISE_976_AES`	15,632	15,624	24
`IQR_SAMWISE_976_CHACHA20`	15,632	15,624	24

SIDH Keys and Secrets

SIDH key and secret sizes:

Table 12. SIDH Sizes
Variant	Public Key (bytes)	Shared Secret (bytes)
`IQR_SIDH_P434`	330	110
`IQR_SIDH_P503`	378	126
`IQR_SIDH_P610`	462	154
`IQR_SIDH_P751`	564	188

Classic McEliece Keys, Ciphertexts, and Secrets

Classic McEliece key encapsulation mechanism key, ciphertext, and shared secret sizes:

Table 13. Classic McEliece Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Ciphertext (bytes)	Shared Secret (bytes)
`IQR_CLASSICMCELIECE_6`	1,047,319	738,218	226	32
`IQR_CLASSICMCELIECE_8`	1,357,824	853,832	240	32

FrodoKEM Keys, Ciphertexts, and Secrets

FrodoKEM key, ciphertext, and shared secret sizes:

Table 14. FrodoKEM Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Ciphertext (bytes)	Shared Secret (bytes)
`IQR_FRODOKEM_640_AES`	9,616	19,880	9,720	16
`IQR_FRODOKEM_640_SHAKE`	9,616	19,880	9,720	16
`IQR_FRODOKEM_976_AES`	15,632	31,280	15,744	24
`IQR_FRODOKEM_976_SHAKE`	15,632	31,280	15,744	24
`IQR_FRODOKEM_1344_AES`	21,520	43,064	21,632	32
`IQR_FRODOKEM_1344_SHAKE`	21,520	43,064	21,632	32

Kyber Keys, Ciphertexts, and Secrets

Kyber key encapsulation mechanism key, ciphertext, and shared secret sizes:

Table 15. Kyber Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Ciphertext (bytes)	Shared Secret (bytes)
`IQR_KYBER_512`	800	1,864	736	32
`IQR_KYBER_768`	1,184	2,760	1,088	32
`IQR_KYBER_1024`	1,568	3,656	1,568	32

NTRUPrime Keys, Ciphertexts, and Secrets

NTRUPrime key, ciphertext, and shared secret sizes:

Table 16. NTRUPrime Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Ciphertext (bytes)	Shared Secret (bytes)
`IQR_SNTRUP_653`	994	1,526	897	32
`IQR_SNTRUP_761`	1,158	1,771	1,039	32
`IQR_SNTRUP_857`	1,322	2,007	1,184	32
`IQR_SNTRUP_953`	1,505	2,262	1,349	32
`IQR_SNTRUP_1013`	1,623	2,425	1,455	32
`IQR_SNTRUP_1277`	2,067	3,067	1,847	32

Saber KEM Keys, Ciphertexts, and Secrets

Saber key, ciphertext, and shared secret sizes:

Table 17. Saber Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Ciphertext (bytes)	Shared Secret (bytes)
`IQR_LIGHT_SABER`	672	1,736	736	32
`IQR_SABER`	992	2,568	1,088	32
`IQR_FIRE_SABER`	1,312	3,400	1,472	32

SIKE KEM Keys, Ciphertexts, and Secrets

SIKE key, ciphertext, and shared secret sizes:

Table 18. SIKE Sizes
Variant	Public Key (bytes)	Private Key (bytes)	Ciphertext (bytes)	Shared Secret (bytes)
`IQR_SIKE_P434`	330	382	346	16
`IQR_SIKE_P503`	378	442	402	24
`IQR_SIKE_P610`	462	532	486	24
`IQR_SIKE_P751`	564	652	596	32

Build Options

During development of the toolkit, builds use many compiler and linker flags intended to help reduce errors and improve overall code quality.

The toolkit libraries are generally built using the "native" compiler (clang or gcc) for a platform:

Table 19. Platform Compilers
Platform	Compiler
Android	`clang`
FreeBSD	`clang`
iOS	`clang`
Linux	`gcc`
macOS	`clang`
Windows	`gcc`

Compiler Options

When building with clang:

-Weverything -Wno-deprecated -Wno-deprecated-declarations -Wno-vla -Wno-packed -Wno-padded -Wno-disabled-macro-expansion -Wno-documentation-unknown-command -Wno-missing-field-initializers -Werror
-Winline (for everything except Android)
-Wno-reserved-id-macro (for everything except FreeBSD)
-fno-stack-protector -fvisibility=hidden -fPIC
-std=c99
-O3
-DNDEBUG -D_FORTIFY_SOURCE=2
-D__USE_MINGW_ANSI_STDIO=1 (Windows only)

When building with gcc:

-Wall -Wextra -Waggregate-return -Wbad-function-cast -Wcast-align -Wcast-qual -Wfloat-equal -Wformat-security -Wformat=2 -Winit-self -Wmissing-include-dirs -Wmissing-noreturn -Wmissing-prototypes -Wnested-externs -Wno-deprecated -Wno-deprecated-delcarations -Wold-style-definition -Wpedantic -Wredundant-decls -Wshadow -Wstrict-prototypes -Wswitch-default -Wuninitialized -Wunreachable-code -Wunused -Wvarargs -Wwrite-strings -Werror
-Winline (for everything except Android)
-fstrict-aliasing -fstrict-overflow -funsafe-loop-optimizations -ffunction-sections -fno-stack-protector -fno-dse -fPIC
-pedantic
-pipe
-std=c99
-O3
-DNDEBUG -D_FORTIFY_SOURCE=2
-D__USE_MINGW_ANSI_STDIO=1 (Windows only)

Linker Options

When building with clang:

-Wl,-undefined,error (for macOS) or -Wl,--no-undefined (others)
-Wl,-unexported_symbol,_isc* and -Wl,-unexported_symbol,_ISC* (for macOS)

When building with gcc:

-Wl,-dead_strip (for macOS) or -Wl,--gc-sections (others)
-Wl,-undefined,error (for macOS) or -Wl,--no-undefined (others)
-Wl,-unexported_symbol,_isc* and -Wl,-unexported_symbol,_ISC* (for macOS)

Apple Bitcode

The toolkit does not support Apple’s Bitcode for iOS, tvOS, or watchOS binaries. For security reasons, the toolkit needs fine control over optimizations and the life cycle of memory buffers, which isn’t possible when producing Bitcode.

Deprecated APIs

As standards for quantum-safe algorithms evolve, the toolkit’s APIs will change to reflect these changes.

If an API is likely to change in the next version of the toolkit, it’ll be marked with IQR_DEPRECATED in the header. For example:

// From iqr_kdf.h:

#ifndef IQR_IGNORE_1_5_DEPRECATED
IQR_DEPRECATED_MSG("The KDF API will be changing in the next toolkit release.")
#endif
IQR_API
iqr_retval iqr_RFC5869HKDFDeriveKey(const iqr_Context *ctx,
    iqr_HashAlgorithmType hash_algo, const uint8_t *salt, size_t salt_size,
    const uint8_t *ikm, size_t ikm_size, const uint8_t *info, size_t info_size,
    uint8_t *key, size_t key_size);

Compiling code that uses iqr_kdf.h will generate a warning (or error if you’re building with -Werror) so you can plan for the upcoming change.

After you’ve created a task (or user story, or bug) to update the code when the next version of the toolkit is released, you can add -DIQR_IGNORE_1_5_DEPRECATED to your compiler flags. This will suppress the warnings for deprecations in version 1.5 of the toolkit.

Code Stripping

The toolkit has been designed to maximize code stripping to help when deploying to embedded systems. This is the reasoning behind having no pre-registered default for the hash algorithms. If you provide your own implementation, or only use a subset of the toolkit’s implementations, your executable won’t include the unused toolkit code.

Internally, algorithms with variants (such as Dilithium) use the same technique to include only the code required to implement the specific variants you use in your applications.

If your compiler/linker supports it, use the -flto option to enable full link-time optimization.

Linux libc

The Linux version of the toolkit is built and tested against the following versions of the libc library:

2.23 (Ubuntu, 64 bit)
2.24 (Arch, 64 bit)
2.24 (Raspbian, 32 bit)

Building on Windows

When building an application that links against the toolkit static library (libiqr_toolkit_static.lib), add the library under Linker > Input > Additional Dependencies in Visual Studio. No other changes to your Solution file are necessary.

Adding a static library to the solution

When linking dynamically, you must specify the import library (libiqr_toolkit.lib) under Linker > Input > Additional Dependencies. Additionally, you must add IQR_DLL to the preprocessor listing under C/C++ > Preprocessor > Preprocessor Definitions. This will ensure that extern symbols are imported correctly. The toolkit DLL (libiqr_toolkit.dll) must reside in a location where it can be found by the linker.

Adding IQR_DLL to preprocessor definitions

Samples

The samples include a Visual Studio 2017 solution (targeting build tools v141 and the Windows SDK version 10.0.17763.0) and project files. These have relative references to the include and lib directories in the toolkit’s main directory.

Legal

The ISARA Radiate Quantum-Safe Library is licensed for use:

The code and other content set out herein is not in the public domain, is considered a trade secret and is confidential to ISARA Corporation. Use, reproduction or distribution, in whole or in part, of such code or other content is strictly prohibited except by express written permission of ISARA Corporation. Please contact ISARA Corporation at info@isara.com for more information.

Please refer to your sales/support contract for more information about technical support and upgrade entitlements.

Trademarks

ISARA Radiate™ is a trademark of ISARA Corporation.

Sample Code License

Sample code (and only the sample code) is covered by the Apache 2.0 license:

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

Apache License 2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Patent Information

Portions of this software are covered by US Patents 9,614,668, 9,660,978, 9,673,977, 9,698,986, 9,780,948, 9,794,249, 9,912,479, 9,942,039, 9,942,040, 10,031,795, 10,061,636, 10,097,351, 10,103,886, 10,116,443, 10,116,450, 10,218,494, 10,218,504, 10,313,124, 10,404,458, 10,425,401, 10,454,681, and 10,581,616.

Portions of this software are covered by Japanese Patent 6,644,894.