CSA Brute Force

dale_para_bajo

Registered
Messages
646
@kewbien, That is the reason I do all for my old AMD HD7950.

I still need to check on those details. You said
"1 chip running 32 cores @200Mhz clk will do a complete search in about 12Hours ... using about 15 Watt" and you said "Dev boards goes easy + $1000 but typical"

So is this typical dev board for $1000 be able to do complete search in about 12Hours? So we can conclude $1000 more or less for that capability? People do not forget the LONG time learning curve.

@Stefan2k16 nicde resume I will keep it for future use.


And
 

dale_para_bajo

Registered
Messages
646
Lets be more descriptive. Can we say that this evaluation board can do 12Hr Full range

Altera Cyclone V E FPGA Development Kit

Click to enlarge

I see many low cost EB from, $99.
What is the spec of the main chip that we required?

This is an example of EB boards
Code:
http://www.terasic.com.tw/cgi-bin/page/archive.pl?**********English&CategoryNo=13&List=Simple

*** characters are "Language" + "=" forum blocks that
 
Last edited:

dale_para_bajo

Registered
Messages
646
I wish I could edit my post so that I do not have to send so many post.

Your comment
"I use Altera Cyclone V E A9
1 chip running 32 cores @200Mhz clk will do a complete search in about 12Hours ... using about 15 Watt"


1) Evaluation Board. It does not comes in A9 it is an A7.

Now looking at DataSheet
Click to Enlarge

1)It shows that A7 has about half of the resources of an A9.
2) Cores: I can not find reference about cores. Much I found is Logic Elements. A7 has 149.7K and A9 301K. So how this Logic Elements relate to Cores?

I see that this FPGA are in fact mutant as they have A9 ARM dual-core CPU. Interesting. Sorry that I am posting a lot but not many people here offer help in FPGA. SO I guess I am trying to get as much knowledge that I can from you. Please receive my thanks ahead.
 

K2TSET

Registered
Messages
125
1) Evaluation Board. It does not comes in A9 it is an A7.

1)It shows that A7 has about half of the resources of an A9.
2) Cores: I can not find reference about cores. Much I found is Logic Elements. A7 has 149.7K and A9 301K. So how this Logic Elements relate to Cores?

I see that this FPGA are in fact mutant as they have A9 ARM dual-core CPU. Interesting. Sorry that I am posting a lot but not many people here offer help in FPGA. SO I guess I am trying to get as much knowledge that I can from you. Please receive my thanks ahead.

Pretty expensive, but a good board

https://www.altera.com/products/boards_and_kits/dev-kits/altera/kit-cyclone-v-gt.html

This one was very low price but also a slower speed grade fpga but still a A9, unfortunately it is obsolete as far I know

http://www.alterawiki.com/wiki/BeMicro_CV_A9#Example_Designs

Correct A9 is double of A7

When I say 1 core it is a complete BF so 32 cores are 32 times BF on a A9 (so A7 would be 16 cores) and so on ... if you go for like a Cyclone 3 or 4 and get some of the biggest (120) you can have about 8 cores.

You still need like some additional stuff besides the BF cores like some interface so send the start data and 1 BF core to validate a second set of data and some output to see what happens on the board.

I would for sure say if you want learn go for a smaller chip like the MAX10 M50 and learn the FPGA stuff.... than later go for a bigger board.

For 1 BF core you will need 28 units Block memory M9K or M10K since they can be split into 2 x ROM each to have the 56 step of BC
and a lot of LE's but you can do the math when you know I can have 32+1 cores in a A9 with 301k LE's and 1220 M10K

Regarding the versions with build in ARM ... sure it nice, but you can use a CPU core like the NIOS if you need it.

But if you just want to learn you can also use the dev software quartus without any board to see how much resources your design will use and even simulate the design, but for sure it will not get any keys :)
 

dale_para_bajo

Registered
Messages
646
I really appreciate you time and effort explaining me this stuff.

When I say I am broke I am not joking. But I do like to learn. Less that I can do is understand FPGA Solution and its requirements. Clearly I can not play. Well I will check on that "dev software quartus".

Now You did not answer me CLEARLY my question. See many here (Including me) do as parrots. We READ and REPEAT time after time same thing. But thruth is that we know nothing. So FPGA?, we people respond just as kebien said: Prepare to spent LOTSS of money.

So here we are and I need this question answer. SO PLEASEEEEEEE be more specific.

Please define better what you call 32 cores?
Do ALL Altera Cyclone V E A9 have 32 cores? I mean 1 single chip as in on the evaluation board. So that we can figure out the cost for 1 day BF.

Is 1 single (Altera Cyclone V E A9 ) = 32 cores = 1 day BF full range.
This is important also. As many here see the FPGA Solution as a Modular System holding many chips or board with chips. But in fact 1 single chip in 1 board can do the job???

It seems I need to find a LOW cost EB with Altera Cyclone V E A9.

Let me also establish my line of thought. I know many will go for the lower cost EV so that they can start writing routines. Well that may work for many. But I do think is a waist of time. You will spend many hours of learning just to find your self time after time in a failure as you need more resources. Then when times comes to move to a new system you will find your self learning all over as the new system is different. So I am not ignoring your offer to go for a system like MAX10. It is just I had see this situation. At the moment I want to know what works, Then I start reading on that particular solution.

So Is 1 single (Altera Cyclone V E A9 ) = 32 cores = 1 day BF full range.
 

Stefan2k16

Registered
Messages
44
So, you've determined to go the FPGA route? There's no doubt an FPGA will probably give you the absolute best performance and will do so at less power consumption and heat output. However, I think it's going to require as much, if not greater, initial investment.

As I mentioned in my earlier comment, cudabiss was written to allow you to run multiple instances of it and each instance can run on different GPU's. All you have to do is have different folders with their own copy of cudabiss, use the command line argument in your shortcut to each instance to select different cards, and have the "input.txt" file specify a different search range for each copy of cudabiss. This opens the door to scalability by simply adding more cards. PCIe bandwidth does not have a big impact on the performance of cudabiss like it might gaming. Therefore you could even use x1 PCIe slots.

I agree with Stefan2k16, CudaBiss is an old program made for early CUDA boards and it might be better with 4 times old card then a newer one.

well, what I really meant there was that using multiple lower performance cards might have advantages with regard to cost, dealing with heat output and power delivery, and the ability to easily scale by simply adding more cards as you can afford it. The top of line cards are the bleeding edge. they are the technology pushed close to the limit. As such they carry a premium price tag. Also I suspect that a newer better optimized cudabiss clone could probably get better performance from newer cards. Though Maxwell and Pascal do run cudabiss quite well despite the fact it's an old piece of software and probably not as well optimized for them. It was Kepler that seemed to do poorly. One card that did run cudabiss well was the GTX 750 Ti. These were 60 watt <$150 cards that could perform like a GTX 580 on cudabiss. Since they were only 60 watt cards, you could have 4 of them in one system and the added power and cooling requirements would be no more than that of one top of the line card which are typically in the 200-250 watt range. And there is the advantage of not having all that heat packed in one slot. Anyway, right now, I would probably look at the GTX 1050. This looks like an interesting card. It has the same number of cuda cores as a GTX 750 Ti, but they are clocked higher and are the newer Pascal architecture therefore should deliver better performance. And if you shop around, you can find them around $100. At that price you could buy 8 of them for 50% less than the cost of your FPGA development board.
 

dale_para_bajo

Registered
Messages
646
@Stefan2k16

I like your explanation. I really do. Well as I said I read everyone comments on this forum and try to understand his view. For the most part I agree with you on GPU. Well I do like OpenCL. Never done CUDA.

Now Just try to see my point. At the moment I am NOT Buy anything nor going the FPGA Route. I am broke. No money.

But that does not mean I can not learn. In the other hands FPGA is like a Myth, I already said that many here including me thought FPGAis REALLY expensive. I could make the assumption that many here think that a FPGA solution require many $10,000 of USD. Well I guess that was true 8 to 10 years ago. But I have a felling new devices are more like in the 1,000 range. And that Is what I try to establish here. Cost. The other what device is capable of providing such a solution.

Now If I get help pointing me out in a good direction, at least I can assure you I will do try to learn what I can with what ever resources I can put my hand on.

So I will still like to know the answer of my last question of my previous post.
 

K2TSET

Registered
Messages
125
Please define better what you call 32 cores?

Do ALL Altera Cyclone V E A9 have 32 cores? I mean 1 single chip as in on the evaluation board. So that we can figure out the cost for 1 day BF.

Is 1 single (Altera Cyclone V E A9 ) = 32 cores = 1 day BF full range.
This is important also. As many here see the FPGA Solution as a Modular System holding many chips or board with chips. But in fact 1 single chip in 1 board can do the job???

It seems I need to find a LOW cost EB with Altera Cyclone V E A9.

Let me also establish my line of thought. I know many will go for the lower cost EV so that they can start writing routines. Well that may work for many. But I do think is a waist of time. You will spend many hours of learning just to find your self time after time in a failure as you need more resources. Then when times comes to move to a new system you will find your self learning all over as the new system is different. So I am not ignoring your offer to go for a system like MAX10. It is just I had see this situation. At the moment I want to know what works, Then I start reading on that particular solution.

So Is 1 single (Altera Cyclone V E A9 ) = 32 cores = 1 day BF full range.

You will need to see a lot of youtube with "what are FPGA" and "FPGA tutorial" etc.

And please start with some very simple until you understand how you can have logic in a FPGA ... without any real logic :D
It's all small lookup tables where you program truetables of the logic in the LUT (S-RAM)
The FPGA have a huge routing system and a lot of latches (FlipFlops) and the you have the block ram's and some i/o

So back to your questions.
What is 1 core .... it's just like a function in normal programming langues, you define a block with some input and some output (in this case a whole BF) depending what inside this block it will use a number of resources

1 core of BF could have let's say
28 units of M10K BRAM
4,000 ALM
12,000 registers

This 1 core might run @200Mhz as pipeline so you will get an output result on every clk (remember all the registers in the 32 cores are clocked on every clk so the FPGA will actually need a small fan and a good PSU even it's only in the power area, with out you will destroy the FPGA)

1 core of BF gives then 200.000.000 results every sec and it the output = 0x00 0x00 0x001 you have a key which fit...
Sure you need to check this against the 2. set of input data

So Is 1 single (Altera Cyclone V E A9 ) = 32 cores = 1 day BF full range.

NO! It's even better it will do a full search in 12 Hours


32 x 200.000.000 x 60 x 60 x 12= 276,480,000,000,000

2^48 = 281,474,976,710,656

So how many cores can you run in a FPGA.... only depends on the numbers of ALM / BRAM and registers

For the BF in a A9 there can be 32 + 1(for 2. check) cores + some control stuff ... there might have been 1 or 2 more cores in the FPGA ...but 32 are a magic number to spread ranges of search.

If you fill a FPGA 100% it will never be able to place / route the solution and maintain the speed at the same time

If you lets say wanted to do something else in the FPGA like BF MD5 or DES or something else your core would have a different size therefore you might have 100 cores in 1 chip or maybe only 2 cores it all depends on what's in the core, hope it makes sense

How fast?
Depend on the FPGA speed grade and model (faster = more cost)

I choose the A9 due to price / capacity and speed

Will it all happens in 1 chip (compared to dev board with many chip on) Yes it all happens in 1 FPGA chip + eeprom which does hare the code to load into the FPGA after power up (build into the MAX10) + some power supply chip's

For programming you will need a JTAG interface (the bigger dev board typical have it build in) to send the code to the FPGA / eeprom.

It might be interesting to have a look at https://opencores.org/
 

dale_para_bajo

Registered
Messages
646
You guys are amazing. One is trying to get me away of FPGA. The other already started the lessons. HEHEHEHE

Listen I know I will have to learn a lot. Usually read read and then read. Do not worry about that, as I am good at reading.

But I only wanted to know a base FPGA setup that can do BF Full range in 1 Day. Less time well that is better.

Now even when you did not answer 100% clearly as you started with the lesson. I will assume that your 12 hour calculation means yes. 1 Single Altera Cyclone V E A9 will do it. In other words $1000 USD EB can do it. Clearly one of this chips with just minimal resources can do the job as you mention. So maybe $200-$500 can do it. But then more hi level of understanding is required. you mention Jtag, My best guess is having a Arm dual-core could mean it has also some sort of Boot programing.

I hope I understood correctly. In any case be sure I will do my best to start learning. I done all sort of microprocessor programing from motorola, intel, Arm, mips etc. PLay with thueth tables and Karnut Maps etc. But never program FPGA. That will be interesting I guess.
 

K2TSET

Registered
Messages
125
you mention Jtag, My best guess is having a Arm dual-core could mean it has also some sort of Boot programing.

JTAG are a standardized interfacing for programming devices MCU / FPGA etc

In my case only for the FPGA ... I have no ARM inside the FPGA.

I just have a simple RS232 to send data for the BF cores and a VGA output to see what happens
 

dugi

Registered
Messages
17
can someone please help me searching cudabiss ?

000000000000
FFFFFF000000
47420097FBD6514CF8A45710DEDC2BA8
47420090ACA55EC7AD6798DA9939A615
4742009C34C7B45CBFF0718AEDBE387B
1
1
 

dale_para_bajo

Registered
Messages
646
JTAG are a standardized interfacing for programming devices MCU / FPGA etc

In my case only for the FPGA ... I have no ARM inside the FPGA.

I just have a simple RS232 to send data for the BF cores and a VGA output to see what happens

Listen I had spent some time searching for Board that contains " Altera Cyclone V E A9 ". Altera sites seems to be dying! Weird. As they offer links to 404 pages? Well If a look for post 2014 in google they even said that Altera seems to offer things that they do not have in hand. As many waited for month to get a hold on one of those Bemicro boards.

So looking at Fleebay I see a couple of used EB that have Cyclone V. But in general it seems not available or an already old device. Listen I do not have the money so this is really a very low priority. But I will like to search for a similar device that has availability. What is the board PN you have in hand?

It will be easier to understand your comments if I read its specs. But in general yes I did read that version "E" has no Arm CPU. I understand and use JTAGs. Well old stile on Routers and some receivers.

RS232 see you may be using some type of bootloader. Now I do not understand "VGA output" Do you mean your board has a DB16 VGA output that you hook to a monitor?
 
Last edited:

dale_para_bajo

Registered
Messages
646
I spent another day searching for your NICE solution. a Dev Board with 5CEFA9F23C. In general Only I found is Arrow BEMICRO CV A9. But it seems it unavailable. I can find chips but not DB.

Now see my personal dilemma. I am not trying to buy an item. But for my personal choice I will not start learning something that is not or will not be available. Most of ebay stuff is bas on really low cost as learning tools. Altera is now own by Intell. Intel university is base on a few old DB. The most closes is a Ciclone V SE with A4 wish has some where from a 6th to a tenth of the resources of an A9.

Do you have any suggestion?
 

fiji

Member
Messages
1,086
@ dale_para_bajo
what about said china made low cost hd receiver
can auto biss supported
search auto biss in google all kind of verity available on ebay
what is method to use decrypt biss channels ?

here is auto support biss receiver :
https://www.alibaba.com/product-detail/Freesat-v7hd-satellite-receiver-cccam-auto_60652967275.html?s=p
OzFyn2i.jpg
 
Last edited:

kebien

Registered
Messages
1,329
can someone please help me searching cudabiss ?

000000000000
FFFFFF000000
47420097FBD6514CF8A45710DEDC2BA8
47420090ACA55EC7AD6798DA9939A615
4742009C34C7B45CBFF0718AEDBE387B
1
1
Tell us what this is
Feed or channel,date,post all information you have.
Makes no point to search for keys if they are not going to be found or maybe is already found.
Also post the TS this packets came from.
But this is not the thread for this,use the appropiate thread and section.
 
Last edited:

K2TSET

Registered
Messages
125
RS232 see you may be using some type of bootloader. Now I do not understand "VGA output" Do you mean your board has a DB16 VGA output that you hook to a monitor?

There are no build in VGA hardware nor DB15 output, I just use 5 pin as output as RGB and HV, same with RS232 just 2 pins no cpu just implemented as"hardware" in the LUT's. you will find plenty of examples on the web.

It's correct that it can be hard to find dev board, they typical comes when new chips are released and I guess they are for RD and not general use as a product since they are expensive, the goal with dev boards are to have manufacturers testing and the producing their own board with the chips.

I think the 404 you see on som altera pages and due to a huge move to Intel webpages. Xilinx and Altera are the biggest on the FPGA marked and Altera are using Intel factory to produce silicon so they will be here for long time.

You should not be so afraid to use different chips the all work more or less the same way and the languages VHDL or verilog are the same for all, but sure RAM blocks and other special stuff like PLL have to be adjusted to fit the chip as well as the pinout.

Buy a cheap one and start playing... you can do so much with those chips and it great to understad how it works.

This should not be a thread about FPGA but more on new ideas to shortcut the CSA algo... any new ideas?
 

dale_para_bajo

Registered
Messages
646
I am so sorry about the hijack. I was under the impression that you do FPGA and some how wanted more ideas in that area. So I wrongly thought that FPGA talk meet your criteria. I will be posting my question in a new thread so that we can keep CSA BF attack Ideas here.

I will be around if I have more Ideas my mind was busy selecting a devices.
 

C0der

Registered
Messages
267
If the state of the SC is known at some point after the first 32 steps, is it possible to "roll back" to get the CW?
 
Last edited:

K2TSET

Registered
Messages
125
If the state of the SC is known at some point after the first 32 steps, is it possible to "roll back" to get the CW?

If you mean in the init phase then I would say so, but in the run phase the 2 output bit are a xor of 4 bit which open up for many combination.

If you know the SB0 input and you know a stage in the SC it would be very fast to BF all combination

How would you know the value of a stage in the SC?

If you try to go backward from the 0x00 0x00 0x001 the BC can have all combination so SC again might have all combination?

The SC are only a forward algo

But there might be some trick around the where the init phase goes to the Run phase since the SB input are used in 4 rounds so somehow round 28..32 might only affect the output of run phase round 1..4 very little

It might lead to a reduction of the key space, but I haven't found it :confused:
 

K2TSET

Registered
Messages
125


Here you have the SC in a more clear way than on the Wiki
All thick lines are 4 bit thin = 1 bit

It's show the Run phase

In the Init phase it the 4 bit from the combiner D0,D1,D2,D3 goes back to the XOR in FRS1
 
Top