/aes256

A byte-oriented AES-256 implementation.

Primary LanguageCMIT LicenseMIT

A byte-oriented AES-256 implementation

Here is a compact implementation of AES-256 in C. It uses only a byte type and is portable enough to fit even an 8-bit computer. It also has a small memory footprint because S-box operations use on-the-fly calculations instead of lookup tables.

Usage

Add aes256.c and aes256.h to your project and compile as usual.

Define BACK_TO_TABLES if you want to get a faster version that will use lookup tables instead of calculations.

There is also the Makefile. Use make help for details.

Modes

This source code gives you a codebook, an ECB core that works on a 16-byte block, and it is up to you to choose and implement an appropriate block cipher mode yourself.

Check the ctr directory for a counter mode complementary implementation.

Portablility

I try my best to keep this source code portable. But supporting a wide range of esoteric or long-obsolete compilers is no longer a priority. Look at the basic edition if you still need that support and can't use the current variant.

The primary development is with clang and GCC on macOS and Linux platforms.

Correctness

Test Vectors

The demo.c uses a few known-answer tests from several official documents to verify this is a valid AES-256 implementation.

Formal Verification

We rely on C Bounded Model Checker to formally verify code properties.

Use make verify to verify all the aes256-prefixed functions bound with the demo code.

If you want to focus verification on a single function, use make verify FUNC=aes256_XYZ, where aes256_XYZ is a function name.

Check https://github.com/diffblue/cbmc for the latest version of CBMC.

History

I initially wrote this source code in 2007 in response to a request from colleagues for a compact implementation of AES-256 in C.

They didn't worry about performance; they only needed something portable to produce small code without using Assembly language.

Unfortunately, none of the publicly available AES implementations at the time fit the bill.

So, I made this straightforward and a somewhat naïve byte-oriented implementation that was slower and more predisposed to potential side-channel attacks by nature. Still, it did the trick.

Later, in 2009, Hal Finney convinced me to make a few changes so that the implementation could benefit a wider audience. We improved decryption subroutines and added an option to compile a faster version with pre-calculated lookup tables.

Legacy

The original version of the source code from http://www.literatecode.com is available under the tag legacy. It is provided only for reference.

See also

Since its first public release, many people have used this implementation for cool things and exciting projects. Not all of those were public. Also, I wasn't great at tracking. Still, there are a couple that may interest you:

The chances are your project could be interesting. So do not hesitate to share.