CIHANGIR TEZCAN


. Home
. Publications
. Presentations
.
Teaching

 

 




My Google Scholar Page

My DBLP Page
   

SILENT CIPHER

Silent Cipher: Silent Encryption Decryption Algorithm is an AES-like cipher designed in August 2017 during my post doctoral studies at RUHR Universität Bochum with Gregor Leander. It is specifically desgined for GPU architecture.

SILENT directly works on the GPU and do not use any resource of the CPU. Therefore it performs encryption operations on the background without affecting any process the CPU performs, hence the name SILENT.


Security: Using an AES-like structure allows SILENT to achieve the security of AES. However, unlike AES, SILENT does not have any invariant subspace. Thus, it is secure against invariant subspace attacks. Moreover, our design makes SILENT to be more secure than AES against known and chosen-key distinguishers. 


 Performance: SILENT is faster than AES on CPU. However, SILENT is specifically designed for GPU architecture and achieves better results than our best AES implementation (which we believe that our optimizations make our AES implementation to be faster than the publicly known AES implementations for GPU).

The performance comparison of our fastest CUDA implementations of ciphers SILENT, AES, DES, and PRESENT on a single GTX 970 GPU is provided in Figure 1.

 

 Silent Benchmark
Figure 1: Performance comparison of our fastest CUDA implementations of ciphers SILENT, AES, DES and PRESENT on a single GTX 970 GPU.

 

Note that SILENT is approximately 2.6 times faster than AES. Namely, the number of block encryptions SILENT can perform is 2.6 times of AES. Since SILENT has a block size of 64 bits and AES has a block size of 128 bits, this 2.6 times speed-up also pays in encrypting a single big file. For example, our AES implementation can encrypt a file of size 16.92 GB in a second in Counter mode. However, we can encrypt a file of size 22.18 GB in a second in Counter mode using the SILENT cipher.


Specifications: SILENT is a family of ciphers and the parameter selection for each version is not finalized yet. The full specifications and source code will be available soon.

 

Visited Times