AES on CUDA?

Is it possible to implement AES Encryption/Decryption based on CUDA?
Can somebody provide some clue about it?
Thanks!

Google is your friend.
[url=“http://www.manavski.com/downloads/PID505889.pdf”]http://www.manavski.com/downloads/PID505889.pdf[/url]

Thank you very much for your information.

However, I can’t download it.

Could anyone download it?

Thanks!

There’s also a GPU implementation featured in GPU Gems 3.
And the download for the paper works just fine here.

Hey

do you have Source Code about Implementation of AES with CUDA?
Thank you

Could I ask if this is just a curiosity or you want to use it?

I really doubt the practicality of AES and H.264 decoding on GPUs. Sure, it probably will be slightly faster than an ASIC, but the input sizes would need to be huge to achieve this speedup and it will consume 100 watts while the ASIC probably consumes only a few watts. The latency would also be poor compared to the ASIC because you need to give the GPU a huge window of data by buffering.

GPU H.264 encoding would be a good idea because it involves a lot more computation (searching for matching blocks from reference frames). It seems most hardware encoders have many limitations like only reference frame, no 8x8 DCT transforms.

Can anyone provide CUDA code for simple AES encryption and decryption of large file

Can anyone provide CUDA code for simple AES encryption and decryption of large file

Decrypting a single data stream is not going to provide enough parallelism to be worthwhile on the GPU.

The transfer over PCIe bus is going to take longer than the CPU simply accessing its main memory (assuming the file is already loaded to CPU memory).

CPUs with AES NI instruction set (like recent AMD offerings, providing a 400% performance gain over non AES NI implementations) are likely going to outpace a GPU implementation

If you look here, the data rates in MBytes/sec per CPU core are given for several CPUs
https://calomel.org/aesni_ssl_performance.html

You might already exceed the available PCI express bandwidth by using just a few CPU cores of a Ryzen 7 CPU, for example.

GPUs have a clear advantage for applications like

a) password cracking (brute force hashing of a lot of password combinations)
b) computing many hashes in parallel to find hashes matching some criteria (e.g. as done crypto currency mining)
c) finding hash collisions (e.g. for forging message signatures, or “security research”)

…anything cryptographic that doesn’t involve a lot of data transfers really.

Guys, i am doing poject on aes implementation using T-table on CUDA, can someone provide me the code of it.

Google really helps

here is some AES code for CUDA

EDIT: after further review of this code I find it to be normal C++ code renamed to .cu file extensions.