eap Mailing List: Re: EAP-AKA Key derivation

[Thomas Otto (t.otto]

Hi all,

Just for curiosity, I looked at implementations of the IKEv2 PRF and DSS-PRF.
EAP-FAST uses the IKEv2 PRF.
EAP-SIM and AKA uses the DSS-PRF.

The opensource project wpa_supplicant has implemented these KDFs. 
My feeling has been that the DSS-PRF is not as efficient ... however, at least
for the particular implementation of wpa_supplicant, this is NOT true.

The bottom-line (for the urgent reader): 
==> The DSS-PRF is about 7-8x faster (regardless if 140 or 512 byte output).




For the interested reader:
The machine tested is an Amilo notebook, 1024 MB RAM, 1.7 GHz Intel Pentium M.


---------------------------
 parameters: 140 outputsize / byte 
 iterations: 100000 

Testing EAP-SIM PRF ...  100000 iterations: 1.783396 seconds 
Testing EAP-FAST PRF ...   100000 iterations: 12.632675 seconds 

The ratio IKEv2-KDF / DSS-KDF = 7.083494
-------------------
 parameters: 512 outputsize / byte 
 iterations: 100000 

Testing EAP-SIM PRF ...  100000 iterations: 5.855131 seconds 
Testing EAP-FAST PRF ...  100000 iterations: 47.793291 seconds 
The ratio IKEv2-KDF / DSS-KDF = 8.162634-------


The source file is available here: 
http://www.tu-bs.de/~y0013790/prftest.c

This file must be copied into the wpa_supplicant directory, 
it can be compiled with

$  gcc  -MMD -O2 -Wall -g  -I. -I../utils -I../hostapd prftest.c sha1.c md5.c 
common.c -o prftest.o