05-21-2012, 06:19 PM
Thanks to all participants!
When the poll was still running I played around a bit with some of the expected winners. It was very unexpected for me so many people were interessted in the length > 16 support. I quickly did some hacks: We will see performance drops between 50% to 500%, depending on the algorithm, the attack mode and total max length supported. Also the memory requirements increments massive. For example for a single hd6990 cracking MD5 of length 50 took around 20gb host memory. That is all because of the dictionary cache which is required for rule based attacks. In theory it should be possible to code different structures for > 16 but then not for support of rule based attack.
I have to admit, seeing people not interessted in markov really astonished me. Coding for distribution should be more simple. I think we will start with support for this on -lite to get some experience. All I know yet is that the communication protocol should be based on http so that it can go through http proxies.
The minor thing, the increment mode for -plus is already done in beta. That was easy. I've decided to add another minor thing, the hotkey prompt to skip to the next dictionary.
The last thing, adding the TC algorithm, was clearly the most uninteressting thing I can imagine. I am a bit disappointed on this result. Honestly, if I do file based crypt I know I have to choose a very complicated password.
The 2nd placed $6$ however is more interessting. I will give this more priority since this is lower hanging fruit.
Again,
Thanks to all participants !!!
When the poll was still running I played around a bit with some of the expected winners. It was very unexpected for me so many people were interessted in the length > 16 support. I quickly did some hacks: We will see performance drops between 50% to 500%, depending on the algorithm, the attack mode and total max length supported. Also the memory requirements increments massive. For example for a single hd6990 cracking MD5 of length 50 took around 20gb host memory. That is all because of the dictionary cache which is required for rule based attacks. In theory it should be possible to code different structures for > 16 but then not for support of rule based attack.
I have to admit, seeing people not interessted in markov really astonished me. Coding for distribution should be more simple. I think we will start with support for this on -lite to get some experience. All I know yet is that the communication protocol should be based on http so that it can go through http proxies.
The minor thing, the increment mode for -plus is already done in beta. That was easy. I've decided to add another minor thing, the hotkey prompt to skip to the next dictionary.
The last thing, adding the TC algorithm, was clearly the most uninteressting thing I can imagine. I am a bit disappointed on this result. Honestly, if I do file based crypt I know I have to choose a very complicated password.
The 2nd placed $6$ however is more interessting. I will give this more priority since this is lower hanging fruit.
Again,
Thanks to all participants !!!