Last updated on October 17, 2020 by Dan Nanni
Intel Advanced Encryption Standard New Instructions (AES-NI) is a special instruction set for x86 processors, which is designed to accelerate the execution of AES algorithms. AES-based symmetric encryption is widely used in a variety of security applications and protocol implementations (e.g., IPSec, SSL/TLS, HTTPS, SSH). OpenSSL crypto library supports AES-based ciphers as well.
To support available hardware extensions, OpenSSL provides so-called EVP crypto APIs (e.g., EVP_Decrypt/EVP_Encrypt) which can automatically leverage hardware acceleration like AES-NI (if available) and fall back to software implementation (if not available), via a single interface. If you want to check whether currently installed OpenSSL supports AES-NI hardware acceleration, you can test using OpenSSL's EVP APIs.
Before proceeding, first verify that current CPUs have the AES instruction set. For this you can inspect CPU flags as follows.
$ grep -m1 -o aes /proc/cpuinfo
If the output shows
aes, that means AES-NI engine is available on current CPUs.
To check whether OpenSSL can leverage AES instruction sets, you can use OpenSSL's EVP APIs. When EVP APIs are called, they can automatically detect the presence of AES-NI and accelerate AES encryption computations using AES instruction sets. Thus you can compare AES performance with or without EVP functions. If AES-NI is available for OpenSSL, you will see significant performance boost when EVP functions are used.
Let's use OpenSSL's built-in speed test.
To measure AES algorithm speed without AES-NI acceleration:
$ openssl speed -elapsed aes-128-cbc
To measure AES algorithm speed with AES-NI acceleration (via EVP APIs):
$ openssl speed -elapsed -evp aes-128-cbc
The above two example outputs show encryption rates for different block sizes. You can see that AES speed with AES-NI acceleration is about five times higher than non-acceleration. This confirms that AES-NI is enabled for OpenSSL. If OpenSSL cannot leverage AES-NI for any reason, two outputs would show the same performance.
Please note that this article is published by Xmodulo.com under a Creative Commons Attribution-ShareAlike 3.0 Unported License. If you would like to use the whole or any part of this article, you need to cite this web page at Xmodulo.com as the original source.