VAMPIRE	eBACS: ECRYPT Benchmarking of Cryptographic Systems	ECRYPT II

Measurements of authenticated ciphers

[Page version: 20250503 16:11:24]

eBAEAD (ECRYPT Benchmarking of Authenticated Ciphers) is a project to measure the performance of authenticated ciphers. This page presents benchmark results collected in eBAEAD:

• "1536+0 encrypt": Speed (cycles/byte) of encrypting a 1536-byte message with no additional data.
• "1536+0 decrypt": Speed of decrypting a 1536-byte message.
• "1536+0 forgery": Speed of recognizing a 1536-byte forged message.
• "1536+1536": Same but using a 1536-byte message and 1536 bytes of additional data.
• "0+1536": Same but using an empty message and 1536 bytes of additional data. Nothing is encrypted but the additional data is authenticated.
• "64": 64-byte messages (and/or 64-byte additional data).
• "long": long messages (and/or long additional data). Speed is computed as 1/1920 of the difference in cycle counts between handling a 2048-byte message and handling a 128-byte message.

Each table row lists the first quartile of many speed measurements, the median of many speed measurements, the third quartile of many speed measurements, and the name of the primitive. Measurements with large variance are indicated in red with question marks. The symbol T: (starting with supercop-20200816) means that the SUPERCOP database at the time of benchmarking did not list constant time as a goal for this implementation. The symbol T!!! means that constant time was listed as a goal for this implementation, but that the implementation failed TIMECOP. (TIMECOP failures are not necessarily security issues; they can sometimes be resolved by, e.g., declaring that a rejection-sampling condition is safe to declassify.)

There is a separate page with more information about each cipher and each implementation. Designers and implementors interested in submitting new authenticated ciphers and new implementations should read the call for submissions.

Version