brute force password cracking capacity of a CPU

[xxTurbonium]

I am researching things like password cracking and the like, and I'm wondering if there is any way to calculate or predict, with reasonable accuracy, how many "passwords per second" that a given CPU can churn through?

Say a Pentium 4 at 3.2GHz.
vs.
Core 2 Duo 1.6GHz
vs.
dual Xeons @ 3GHz

etc.

I am assuming things like password length are relevant, but I have no idea how you would factor that into a mathematical estimate. How about things like disk and memory speed? Are these even involved in brute force password generation+comparison?

I am basing my "speed measure" of passwords/second on this page, so I'm assuming it's a common measure of speed.

Moved to CPU forum - Moderator Rubycon

 

[Modelworks]

There really are too many variables to be able to make a blanket statement on all passwords.
Memory speed would be the biggest thing next to cpu speed.
If you can fit the entire code into memory you have no need for disk reading/writing.
I would target one form of encryption and focus on that.

That page cannot do the stronger forms of encryption.
One I use has 256 bit passwords that are double encrypted.
If you crack the first encryption key, you still cannot crack the second because it uses information to reconstruct the data that isn't in the first key.

 

[Cogman]

There is no way to say definitively that CPU x will crack a y length password in z seconds. The best you could do is calculate the number of guesses that would be made (if you already know the password)

Things that screw up the calculations

Structure of the bruteforce. Perhaps you us many floating point conversions (for whatever reason) which gives one CPU a benefit over another.

Instruction set differences. Perhaps one cpu has the SSSSSE10 instruction set which was specially designed force password crackers. That would through many things out of whack.

Who is using the computer. If you have a process that starts once every third week of march every leap year (which just so happens to be the day you run the test) then your results will be off because some processing power is going to appx

You could use big O notation, but really all brute force methods should be O(n) where n is the number of tries.

 

[xxTurbonium]

Originally posted by: Cogman
There is no way to say definitively that CPU x will crack a y length password in z seconds. The best you could do is calculate the number of guesses that would be made (if you already know the password)

I'm not concerned about the time it takes to crack something. Just how many possibilities a given CPU can go through per unit time.

I'm not doing anything too fancy. I'm just experimenting with WPA cracking, which to my knowledge uses dictionary attacks to test out possibilities. I'm doing a test run on a 1.6GHz Core 2 Duo CPU, and after 150 minutes, it's gone through about 1.9 million possible keys (which I assume means 1.9 million dictionary words). It also reports an average rate of 205k/s, but I have no idea what exactly this is in reference to.

I'm not aware at the moment as to what each of these key tests entails, but I'm assuming there's a layer or two of math behind each, which leads back to the whole argument of the method being used.

I guess you would have to know the exact process being used to make any sort of estimation on the rate of a brute force attack.

 

[Cogman]

Originally posted by: databird

Originally posted by: Cogman
There is no way to say definitively that CPU x will crack a y length password in z seconds. The best you could do is calculate the number of guesses that would be made (if you already know the password)

I'm not concerned about the time it takes to crack something. Just how many possibilities a given CPU can go through per unit time.

I'm not doing anything too fancy. I'm just experimenting with WPA cracking, which to my knowledge uses dictionary attacks to test out possibilities. I'm doing a test run on a 1.6GHz Core 2 Duo CPU, and after 150 minutes, it's gone through about 1.9 million possible keys (which I assume means 1.9 million dictionary words). It also reports an average rate of 205k/s, but I have no idea what exactly this is in reference to.

I'm not aware at the moment as to what each of these key tests entails, but I'm assuming there's a layer or two of math behind each, which leads back to the whole argument of the method being used.

I guess you would have to know the exact process being used to make any sort of estimation on the rate of a brute force attack.

That would be what we call a contradiction. If you knew how many passwords a cpu could process in a given period of time, then you would know how long it would take a CPU to crack a given answer and vica-versa.

Again, I'll state that it is absolutely impossible to tell how fast a cpu will crack a password. Even identical cpus with identical instruction sets for the reasons given above. The best you could do is say that brute forcing a password should always be in linear time.

If the software starts using tricks like Hash Tables, dictionary attacks, ect, then the realm of how impossible it is to predict speed becomes even greater.

The absolute best you could do is get a website to use the same software to benchmark different CPUs on their WPA cracking rates using various pieces of software. Since the legality of such a review would be in question, I doubt you will find many websites that regularly update data like that.

Given any computer program, it is impossible to predict ahead of time if one cpu will preform better then another. Go look at any benchmark in any review and you will see that in almost all cases there will be a benchmark where one cpu does worse and then better then another, even in the same genre (for example, games) with sometimes huge performance leaps IE going from being 15% slower to 15% faster.

Ok, maybe it would be possible for a super computer to run every singly possible outcome if it had the schematics of the CPU and every possible situation it would be in to predict which CPU you should buy to crack your neighbors WiFi the fastest, but that is far more expensive then just buying every cpu available regularly and testing them out.

In short, No, I can't tell you a sure fire method that will give you the best CPU to purchase so you can go on leaching WiFi from your neighbors.

 

[xxTurbonium]

Originally posted by: Cogman

Originally posted by: databird

Originally posted by: Cogman
There is no way to say definitively that CPU x will crack a y length password in z seconds. The best you could do is calculate the number of guesses that would be made (if you already know the password)

I'm not concerned about the time it takes to crack something. Just how many possibilities a given CPU can go through per unit time.

I'm not doing anything too fancy. I'm just experimenting with WPA cracking, which to my knowledge uses dictionary attacks to test out possibilities. I'm doing a test run on a 1.6GHz Core 2 Duo CPU, and after 150 minutes, it's gone through about 1.9 million possible keys (which I assume means 1.9 million dictionary words). It also reports an average rate of 205k/s, but I have no idea what exactly this is in reference to.

I'm not aware at the moment as to what each of these key tests entails, but I'm assuming there's a layer or two of math behind each, which leads back to the whole argument of the method being used.

I guess you would have to know the exact process being used to make any sort of estimation on the rate of a brute force attack.

That would be what we call a contradiction. If you knew how many passwords a cpu could process in a given period of time, then you would know how long it would take a CPU to crack a given answer and vica-versa.

Again, I'll state that it is absolutely impossible to tell how fast a cpu will crack a password. Even identical cpus with identical instruction sets for the reasons given above. The best you could do is say that brute forcing a password should always be in linear time.

If the software starts using tricks like Hash Tables, dictionary attacks, ect, then the realm of how impossible it is to predict speed becomes even greater.

The absolute best you could do is get a website to use the same software to benchmark different CPUs on their WPA cracking rates using various pieces of software. Since the legality of such a review would be in question, I doubt you will find many websites that regularly update data like that.

Given any computer program, it is impossible to predict ahead of time if one cpu will preform better then another. Go look at any benchmark in any review and you will see that in almost all cases there will be a benchmark where one cpu does worse and then better then another, even in the same genre (for example, games) with sometimes huge performance leaps IE going from being 15% slower to 15% faster.

Ok, maybe it would be possible for a super computer to run every singly possible outcome if it had the schematics of the CPU and every possible situation it would be in to predict which CPU you should buy to crack your neighbors WiFi the fastest, but that is far more expensive then just buying every cpu available regularly and testing them out.

In short, No, I can't tell you a sure fire method that will give you the best CPU to purchase so you can go on leaching WiFi from your neighbors.

That's a very detailed answer for someone so skeptical of my intentions.

Are you going to help without being patronizing, or are you just going to take me for a malicious cracker and ignore me? Make up your mind.

Anyhow, I already stated I'm experimenting with WEP/WPA cracking out of curiosity. And I'm interested in predicting CPU performance in different applications, such as game servers, etc.

And it's not like I'm trying to figure out which CPU is the "best bang for the buck" so I can go buy it to crack people's WPA connections. At that price, I would just get my own Net access.

Honestly... how else am I supposed to experiment with password cracking, dictionary attacks, etc? And I know of no other way to actually toy with WiFi networks and learn how they work.

 

[Evadman]

If you are seriously interested in cracking your own WPA for fun and want to play around, then forget brute forcing. That is ludicrous, and I don't mean ludicrous speed. Play around with rainbow tables (the hash table that Cogman mentioned). You can get your own table generator, and benchmark a bunch of CPU's if you don't trust the research that has been done along these lines. Check out some of the distributed MD5 hash programs that Berkeley has done (I think it was Berkeley though BOINC). You can also learn more by joining one of the crunch teams, say TeAm AnandTech in the distributed computing forum here at AT.

WPA-PSK is a little more fun than a MD5 or lan manager hash in that the SSID is used like a salt to the WPA hash. That means the hash on WPA for a password of 'evadman' is different if the SSID is 'anandtech' or 'supercoolplace'

 

[sdifox]

Originally posted by: databird

Originally posted by: Cogman

Originally posted by: databird

Originally posted by: Cogman
There is no way to say definitively that CPU x will crack a y length password in z seconds. The best you could do is calculate the number of guesses that would be made (if you already know the password)

I'm not concerned about the time it takes to crack something. Just how many possibilities a given CPU can go through per unit time.

I'm not doing anything too fancy. I'm just experimenting with WPA cracking, which to my knowledge uses dictionary attacks to test out possibilities. I'm doing a test run on a 1.6GHz Core 2 Duo CPU, and after 150 minutes, it's gone through about 1.9 million possible keys (which I assume means 1.9 million dictionary words). It also reports an average rate of 205k/s, but I have no idea what exactly this is in reference to.

I'm not aware at the moment as to what each of these key tests entails, but I'm assuming there's a layer or two of math behind each, which leads back to the whole argument of the method being used.

I guess you would have to know the exact process being used to make any sort of estimation on the rate of a brute force attack.

That would be what we call a contradiction. If you knew how many passwords a cpu could process in a given period of time, then you would know how long it would take a CPU to crack a given answer and vica-versa.

Again, I'll state that it is absolutely impossible to tell how fast a cpu will crack a password. Even identical cpus with identical instruction sets for the reasons given above. The best you could do is say that brute forcing a password should always be in linear time.

If the software starts using tricks like Hash Tables, dictionary attacks, ect, then the realm of how impossible it is to predict speed becomes even greater.

The absolute best you could do is get a website to use the same software to benchmark different CPUs on their WPA cracking rates using various pieces of software. Since the legality of such a review would be in question, I doubt you will find many websites that regularly update data like that.

Given any computer program, it is impossible to predict ahead of time if one cpu will preform better then another. Go look at any benchmark in any review and you will see that in almost all cases there will be a benchmark where one cpu does worse and then better then another, even in the same genre (for example, games) with sometimes huge performance leaps IE going from being 15% slower to 15% faster.

Ok, maybe it would be possible for a super computer to run every singly possible outcome if it had the schematics of the CPU and every possible situation it would be in to predict which CPU you should buy to crack your neighbors WiFi the fastest, but that is far more expensive then just buying every cpu available regularly and testing them out.

In short, No, I can't tell you a sure fire method that will give you the best CPU to purchase so you can go on leaching WiFi from your neighbors.

That's a very detailed answer for someone so skeptical of my intentions.

Are you going to help without being patronizing, or are you just going to take me for a malicious cracker and ignore me? Make up your mind.

Anyhow, I already stated I'm experimenting with WEP/WPA cracking out of curiosity. And I'm interested in predicting CPU performance in different applications, such as game servers, etc.

And it's not like I'm trying to figure out which CPU is the "best bang for the buck" so I can go buy it to crack people's WPA connections. At that price, I would just get my own Net access.

Honestly... how else am I supposed to experiment with password cracking, dictionary attacks, etc? And I know of no other way to actually toy with WiFi networks and learn how they work.

You mean like it is too detailed for you? The man explained why it is not a simple linear equation. Unless you have the exact algorithm, there is no way for you to predict performance.

Here is a hint, NSA doesn't run desktop chips.


And you need to study http://en.wikipedia.org/wiki/NP-complete

 

[Modelworks]

Also don't forget that some things that require passwords have provisions like, enter the wrong password 3 times in a row and it disables the ability to enter anything for anywhere from seconds to hours. That could take forever to brute force attack, since every 3 attempts you have to wait , say 10 minutes before you can try again.

 

[covert24]

http://www.youtube.com/watch?v=BH8X8w8a4f4

 

[xxTurbonium]

Originally posted by: sdifox
You mean like it is too detailed for you?

The point is he calls me a criminal who is seeking knowledge for the wrong reasons, and then proceeds to give me some of that knowledge anyway. Reading comprehension please?

Originally posted by: sdifox
Here is a hint, NSA doesn't run desktop chips.

O RLY?

 

[sdifox]

Originally posted by: databird

Originally posted by: sdifox
You mean like it is too detailed for you?

The point is he calls me a criminal who is seeking knowledge for the wrong reasons, and then proceeds to give me some of that knowledge anyway. Reading comprehension please?

Originally posted by: sdifox
Here is a hint, NSA doesn't run desktop chips.

O RLY?

He gives you a long explanation, you focus on one sentence. That is surely the way to try to understand what he is saying. Did you just skip to the last sentence? I won't comment on your reading skills since you will gleefully ignore it and read some other part of the message. Did you go read up on NP-Complete yet? That might help you understand what Cogman said.

I have probably been messing with computers longer than you have been alive.

 

[xxTurbonium]

Originally posted by: sdifox
I have probably been messing with computers longer than you have been alive.

That's the sad part.

Tell me then, oh old and wise one: In all your years, have you not yet learned...

- to not assume? (you're making assumption as to my age, and you assumed I didn't read what Cogman said)
- to follow an argument and to write proper rebuttals?
- to not argue with people over the Internet?

It seems not.

But hey, you know more about computer than I do, so I guess that fixes everything. You win, good on you.

 

[sdifox]

Originally posted by: databird

Originally posted by: sdifox
I have probably been messing with computers longer than you have been alive.

That's the sad part.

Tell me then, oh old and wise one: In all your years, have you not yet learned...

- to not assume? (you're making assumption as to my age, and you assumed I didn't read what Cogman said)
- to follow an argument and to write proper rebuttals?
- to not argue with people over the Internet?

It seems not.

But hey, you know more about computer than I do, so I guess that fixes everything. You win, good on you.

again, you focus on the last sentence as opposed to the entire argument. I give up.

 

[xxTurbonium]

Originally posted by: sdifox

Originally posted by: databird

Originally posted by: sdifox
I have probably been messing with computers longer than you have been alive.

That's the sad part.

Tell me then, oh old and wise one: In all your years, have you not yet learned...

- to not assume? (you're making assumption as to my age, and you assumed I didn't read what Cogman said)
- to follow an argument and to write proper rebuttals?
- to not argue with people over the Internet?

It seems not.

But hey, you know more about computer than I do, so I guess that fixes everything. You win, good on you.

again, you focus on the last sentence as opposed to the entire argument. I give up.

If you actually read and understood my post, you would see that I responded to your entire statement.

Would me editing my quote to reduce emphasis on your final statement (as was deserved) help you understand?

Nice dodges by the way.

 

[Cogman]

Originally posted by: databird

Originally posted by: sdifox

Originally posted by: databird

Originally posted by: sdifox
I have probably been messing with computers longer than you have been alive.

That's the sad part.

Tell me then, oh old and wise one: In all your years, have you not yet learned...

- to not assume? (you're making assumption as to my age, and you assumed I didn't read what Cogman said)
- to follow an argument and to write proper rebuttals?
- to not argue with people over the Internet?

It seems not.

But hey, you know more about computer than I do, so I guess that fixes everything. You win, good on you.

again, you focus on the last sentence as opposed to the entire argument. I give up.

If you actually read and understood my post, you would see that I responded to your entire statement.

Would me editing my quote to reduce emphasis on your final statement (as was deserved) help you understand?

Nice dodges by the way.

What exactly are you looking for here other then to argue with people? Every question you have asked has been answered.

If we sound like we are patronizing you it is because you initial post (the way things are worded) made it sound like you know nothing of what you are talking about. You say you want to do research into password cracking, and yet you don't know a very basic computer programming concept, that it is impossible to predict program speed on different processors.

Yes, I am suspicious of the fact that you specifically mentioned cracking WPA keys as it sounds like you just downloaded a random program and started trying to crack a key, then came here and stated "Can you predict how fast a computer can crack a key".

Sorry if we come off harsh, but you'll find your basic question has been answered. If you have further questions pertaining to the subject feel free to ask. But if you are just trying to further an internet argument, HT is not the place for it.

 

[Net]

yes you can do a rough calculation. but it will depend on the algorithm. you will have to know details of the algorithm.
what is the overall goal?

250k/s this is 250 keys per second

 

[Concillian]

From the sound of it, you shouldn't be worrying how fast a CPU can brute force your wireless encryption, but rather, how fast an SLi setup can:

http://www.itwire.com/content/view/21098/53/1/0/

http://www.scmagazineuk.com/Wi...ection/article/119294/

Global Secure Systems has said that a Russian's firm's use of the latest NVidia graphics cards to accelerate WiFi ?password recovery' times by up to an astonishing 10,000 per cent proves that WiFi's WPA and WPA2 encryption systems are no longer enough to protect wireless data.

10000% speedup is quite a lot faster.

 

[VirtualLarry]

Originally posted by: Concillian
From the sound of it, you shouldn't be worrying how fast a CPU can brute force your wireless encryption, but rather, how fast an SLi setup can:

http://www.itwire.com/content/view/21098/53/1/0/

http://www.scmagazineuk.com/Wi...ection/article/119294/

Global Secure Systems has said that a Russian's firm's use of the latest NVidia graphics cards to accelerate WiFi ?password recovery' times by up to an astonishing 10,000 per cent proves that WiFi's WPA and WPA2 encryption systems are no longer enough to protect wireless data.

10000% speedup is quite a lot faster.

Now THATS news. Does NVidia make an external Tesla box for laptops, that you could use for some extended wardriving? Hmm.

 

[daw123]

Originally posted by: VirtualLarry

Originally posted by: Concillian
From the sound of it, you shouldn't be worrying how fast a CPU can brute force your wireless encryption, but rather, how fast an SLi setup can:

http://www.itwire.com/content/view/21098/53/1/0/

http://www.scmagazineuk.com/Wi...ection/article/119294/

Global Secure Systems has said that a Russian's firm's use of the latest NVidia graphics cards to accelerate WiFi ?password recovery' times by up to an astonishing 10,000 per cent proves that WiFi's WPA and WPA2 encryption systems are no longer enough to protect wireless data.

10000% speedup is quite a lot faster.

Now THATS news. Does NVidia make an external Tesla box for laptops, that you could use for some extended wardriving? Hmm.

Holy sh*t, that is worrying, although it doesn't mention how many Nvidia cards are required in the system; 10, 20, 100, etc?

 

[FalseChristian]

How long do you think it would take take to crack 50,000 key-code combinations with 15 second intervals? Just wondering.

 

[waffleironhead]

one place to look would be distributed.net if you are concerned with keys/cpu they have a comparison database for their rc5 crack efforts

 

[tenax]

uhmmmm..just wondering where you guys live, what your passwords are and do you have any valuable pron on your computer..

i'm only asking as i like to experiment and want to do some research, though