Selfish Mining

csw [1:36 AM]
@cypherblock
If you have done this correctly.

Explain the data pre and post 2016 blocks.

csw [1:38 AM]
@satoshi
That is the correct value.

Where is the working :wink:
 3 replies Last reply 4 days ago View thread

csw [1:39 AM]
@cypherblock
The maximum Gamma based on the actual Bitcoin network is 0.002

That will come later.

Without Gamma, what are the results?

elliotolds [6:04 AM]
@cypherblock sounds correct to me. props for going through the effort and proving it to yourself

csw [6:08 AM]
Now, the issue is that the State diagram is wrong :slightly_smiling_face:

[6:08]
But we will get to that.

[6:09]
First, @cypherblock

What is the distribution without gamma of the discovery over the first week of blocks?

How many are discovered by your 1/3 SM in a day?

csw [6:16 AM]
Fig 1 - State diagram.

State 3 goes to state 2

Explain how this is the case from the State diagram.... (G) when the pool wins from 1 -2 goes to state 0....
Not state 3- state 2...

Did you not see the contradiction?

[6:19]
Let us take Gamma at zero and ask what is the state table that results?

[6:20]
0.1 BTC for the first person (not you sorry @joeldalais ) to the person who posts it... Ending today

joeldalais [6:20 AM]
:stuck_out_tongue:

csw [6:20 AM]
If 12 hours pass, you can do one @joeldalais

joeldalais [6:21 AM]
i can try :slightly_smiling_face:

thatwildcard [8:01 AM]
it is, and now I know what a MOOC is - great idea :slightly_smiling_face:
iang
https://docs.google.com/document/d/1UiMS4Br7LkTIRdEOPjCukOxG6gIdCJTOgXbeOPBHrcE/edit#
Posted in #selfish-miningMay 14th at 5:07 PM

cypherblock [8:29 AM]
I’m doing some trials now with low gamma.


csw [8:38 AM]
Please track the solutions a day :slightly_smiling_face:

cypherblock [8:55 AM]
Low gamma drastically reduces (or eliminates??) the selfish mining advantage in terms of rewarded-blocks per unit time. Right now my simulations are doing more retargeting than real bitcoin network. But I can see if I can normalize it.

csw [8:56 AM]
As stated... Gamma is an assumption.

[8:57]
The max you can in theory get for Gamma is 0.002 and closer to 0.00098
And this is expensive

[8:58]
To gain a 0.00098 Gamma right now will cost more than adding 10% Hash power to the network...

[8:58]
Hence why I was saying that a test of the network is essential and that experiment is needed :wink:

[9:01]
The network distance is so small and the density of vertices so high, that you will need 2.1x10^4 Sybil nodes for each Gamma gain of 0.00001

[9:02]
To obtain a Gamma that radically alters the network will require 2x10^6 nodes.

[9:05]
Amazon gave me a discounted monthly cost for this. $29,280,000.00 USD

[9:07]
Does anyone argue that I *could not* obtain 10% of the network hash rate for less than the Amazon fees?

cypherblock [9:08 AM]
Sybil nodes would not be the most effective way to do this. Probably do something like this describes: https://arxiv.org/abs/1605.07524

csw [9:08 AM]
True, but the SM paper uses a Sybil node.

[9:09]
That method is also wrong - for a completely different reason, but one thing at a time :slightly_smiling_face: (edited)

[9:11]
The mapping of Bitcoin to the AS system is interesting. I will present this towards the end of the year.... too much to do

[9:12]
The attack would slow transactions, but not blocks. So it is not the same as a SM attack and still does not help

[9:14]
Quotes from SM:
_"These virtual miners act as advance sensors by participating in data dissemination, but do not mine new blocks."_

_"By adding enough virtual nodes, the pool operator can thus increase γ"_

_"The virtual miners are managed by the pool, and once they hear of block X, they ignore it and start propagating block P"_

[9:15]
So, we have:

"But a savvy pool operator can perform a sybil attack on honest miners by adding a significant number of zero-power miners to the Bitcoin miner network"_

[9:17]
This is a logical Fallacy.
http://rationalwiki.org/wiki/Appeal_to_probability

This is an extended for of the logical fallacy, *Appeal to probability*

This is the alternate form of the fallacy known as *"the appeal to possibility*. This flawed logic states in effect _"it is possible, therefore it is certain"_.

[9:20]
*Next*

We come to the CORE of the argument.... What the authors NEED to address and fail miserably on.

_*"Because the protocol was believed to reward miners strictly in proportion to the ratio of the overall mining power they control, a miner in a large pool was believed to earn the same revenue as it would in a small pool."_*

Most critically,
_"We show that, above a certain threshold size, the revenue of a selfish pool rises superlinearly with pool size above its revenue with the honest strategy."_

This is *the* thesis. The hypothesis.

[9:21]
So, if you make more orphan blocks and even if you are close to 50%... you still gain under the revenue you would have gained.

[9:21]
50% of 144 blocks a day = 72 blocks

[9:22]
At best, SM can make many orphans and increase the loss to the other party, but at what cost :slightly_smiling_face:

[9:22]
Remember this is a hypothesis on:

*"We show that, above a certain threshold size, the revenue of a selfish pool rises superlinearly with pool size above its revenue with the honest strategy."*

[9:25]
A rational business thinks differently to an academic.
You think in terms of absolute returns.

Above *ITS* (the selfish miners' revenue with the honest strategy." (edited)

[9:27]
*Next*

The authors say this is difficult to detect and stop....

Clearly they have no idea of statistical detection methods.

[9:28]
In 2008, I wrote a paper and a tool to detect the use of Hydan, a stegonographic tool designed to hide information in code by flipping (ADD A) to (Sub -A) in the binary compiled code

[9:28]
https://uk.sans.org/reading-room/whitepapers/stenganography/detecting-hydan-statistical-methods-classifying-hydan-based-stegonagraphy-execut-32839

[9:28]
That was easy and it was 10,000,000,000,000 times harder than checking the release of blocks in the SM strategy (edited)

[9:32]
Think on it...
For  a place where the SM has a lead:

HM releases, SM orphans it in seconds
HM releases, SM orphans it in seconds
HM releases, SM orphans it in seconds
HM releases, SM orphans it in seconds

HM has lead...

HM finds a block after a long period...
    SM releases two and HM is orphaned

This is the simplest - *PUBLICLY* most simple to check real time attack to stop....

[9:33]
Oh... and the state table is wrong....

If you actually check the processes, the state table never releases 2 or more blocks, but the strategy does...

cypherblock [9:34 AM]
yes, detection seems like it would not be that difficult.

csw [9:34 AM]
And this is not even a third of the way into the problems :wink:

[9:35]
@cypherblock
You detect this - it has been 2 days....

Will Honest miners allow it?

Or will they simply block the Selfish pool (and reject all its blocks (and this does not need code changes...

[9:35]
They could fork the defecting pool...

[9:36]
The pool could (at expense) reorg and try again, but at what gain and what costs?

[9:38]
I will dig up and link the paper later, but it was myself who created and wrote the tool to detect *Hidden* TruCrypt volumes....

The process was used by law enforcement - hence TC is not used now.

cypherblock [9:38 AM]
ok, have to eat now. First thing was to validate that IF gamma of say 50% is used, AND ignoring all other problems, does selfish strategy work? Seems like answer is yes. Next, sure we can show that achieving high gamma is difficult/expensive and detection not so hard.

csw [9:38 AM]
THAT. That was difficult

TC hidden volumes, small entropy changes. Hard

[9:39]
I have a paper and data that will kill Gamma

[9:39]
:slightly_smiling_face:

[9:39]
I have a complete node map of the bitcoin network from 2009 to well this year.

[9:40]
We need to anonymise it prior to releasing the data

csw [9:40 AM]
uploaded this image: image.png
Add Comment

csw [9:41 AM]
Oh, we could detect Hydan at an encoding level of 1/15,000

[9:43]
The SM can be detected at the rate of a SM with 0.000,067% of the overall network and up doing it....
And for a small SM, there is a loss :wink:

[9:43]
I am off for now as well... Enough procrastination. I have a Risk paper for my MSc to get submitted :slightly_smiling_face:


tomothy [11:23 AM]
Gonna invite linzheming if that's ok? Im assuming it is?

csw [11:24 AM]
It is ok

tomothy [11:24 AM]
Oh, he's already here.

linzheming [11:25 AM]
I’m not quite follow the concept sorry.

[11:26]
Is there any definition or theory I can read first?

tomothy [11:26 AM]
It's well above me. From my understanding, emir started from a flawed assumptions and therefore study is flawed.

csw [11:49 AM]
Has anyone thought what happens to the "extra hidden blocks" when the difficulty changes?


[11:49]
The ones at a difficulty that is LOWER than the main honest chain...

[11:50]
They do not need to be far ahead...

[11:50]
Try adding it next :slightly_smiling_face:

david [12:01 PM]
joined selfish-mining by invitation from @onchainscaling

cypherblock [12:46 PM]
@csw yes this is an issue, but I don’t think it is a major flaw in SM algo. I can see if I can account for that in my model (reset private branch to 0 on difficulty change or something).

csw [12:48 PM]
@cypherblock If the SM has a lead, they will have a small difficulty adjustment...

[12:48]
Small is still rejected :slightly_smiling_face:

cypherblock [12:49 PM]
the most number of blocks published at once in SM algo is 2.

csw [12:49 PM]
Yes Published

[12:50]
But Difficulty is calculated by them on their hidden blocks (edited)

[12:50]
And this can differ from the main chain

cypherblock [12:51 PM]
presumably SM miner uses main chain difficulty.

csw [12:51 PM]
They do not know it

[12:51]
Remember, they have hidden blocks

cypherblock [12:51 PM]
They can run ‘border’ node or something to get this. Or multiple listener nodes around the world.

csw [12:52 PM]
Again, they have blocks that are ahead of the change

[12:52]
And they are not able to see the future are they?

[12:53]
Block 2016 for the Honest miner is not the same as Block 2016 for the SM

[12:53]
So, 2017 is not the same when it is republished

cypherblock [12:53 PM]
It is a reasonable issue. I have been thinking about it as well.  I’m not sure if the impact will be huge though. Yes they could predict the future, but that is probably unwise.

csw [12:53 PM]
Each small issue erodes it.

[12:54]
It was always a small gain. And each incorrect assumption makes it smaller and smaller

[12:54]
:slightly_smiling_face:

[12:54]
AND

[12:54]
People are slowly learning how a few things may actually work

jpjp [12:56 PM]
joined selfish-mining by invitation from @tomothy

iang [2:05 PM]
:slightly_smiling_face:

csw [2:25 PM]
Also remember this is off revenue. Each point on revenue makes a massive change in profit.

[2:25]
:slightly_smiling_face:

[2:26]
I guess we forgot that

[2:26]
And non profitable = losses

elliotolds [3:16 PM]
@ cypherblock , note that nothing bad happens to the selfish miners private blocks during a difficulty adjustment, because when he releases his own private blocks, the public chain will adopt the difficulty adjustment in the private chain

[3:17]
@cypherblock

csw [3:30 PM]
Will they now...

[3:30]
Maybe you should test it and not assume.

[3:31]
As the majority calls for a higher difficulty, you will find your statement is not the case

[3:31]
The SM releases and is rejected

[3:31]
And needs to align back to the chain.

[3:32]
Even 2 blocks of 2016 can make a small difference.

[3:33]
The SM can mine on a higher difficultly to ensure that this does not occur, but -
1. This is not a part of the strategy
2. It lowers returns at the end.

elliotolds [3:42 PM]
csw you should try to give a concrete example of the difficulty adjustment causing problems for the SM, and you should see your mistake.

the public chain's difficulty adjustment will be roughly equal to the private chain's adjustment, fyi

csw [3:43 PM]
If you have (HM +2), (SM +4)

SM changes at a lower difficulty than the HM.

Roughly is not the same as equal


[3:44]
Have you tested it :slightly_smiling_face:

elliotolds [3:46 PM]
you mean the chain split is at block X, and HM is 2 blocks ahead and SM is 4 blocks ahead?  the SM's chain's difficulty will end up slightly higher in this case. you seem to have the order backwards.

but the main point is that this doesn't matter, because the HM will publish their blocks, the public chain will start using the HM's difficulty, but then the SM will reveal hidden blocks and the public chain will switch over to the SM's difficulty adjustment

[3:46]
give a concrete example, and you'll see your mistake :slightly_smiling_face:. gotta go for a bit

csw [3:47 PM]
@elliotolds
You seem extremely vested in a flawed model...


I wonder why?

[3:47]
And very much not interested in actually testing it,

[3:47]
Real systems, real data :wink:

[3:48]
And you may want to look up and real on the Negative Binomial distribution.

bitcoinsteffen [4:58 PM]
joined selfish-mining by invitation from @bitsko

cypherblock [9:31 PM]
I will see if I can improve my difficulty calcs a bit more. I am using one simplification which could affect outcome. I think @elliotolds is probably right though, but a couple of things I’m going to check.


----- May 22nd -----
travin
[12:55 AM]
`https://pastebin.com/9s1VF16Z`


csw [1:18 AM]
Here is a big clue. Revenue is a time based and not block based function.

cypherblock [6:38 AM]
my simulations calculate reward-blocks/hour, which should be a good indicator if a strategy has  a revenue benefit or not.

csw [6:52 AM]
What do you get in the first week for a 40% SM @cypherblock ?

[6:53]
They will normally get 403.2 blocks
And how many orphans?

[6:54]
Here is a little secret. In this time frame, the SM can still not mine faster than alpha.

cypherblock [6:55 AM]
Yes there may be a ‘transition cost’ when switching initially from honest to selfish, is that what you mean?

csw [6:56 AM]
In part.

[6:56]
But, remember that even if the HM gives up in the first week leaving ONLY the SM, the SM still only gets alpha of the total

[6:57]
So, at best, they can get as much as they would have received.

[6:57]
AT BEST

[6:57]
Forget this notion of relative pools and other  lies used in graphs to mislead.

[6:58]
Some SM blocks are orphaned. Some HM blocks. It may be more HM than SM, BUT and here is the kicker...

[6:59]
If the SM can at alpha = 0.4 earn at most 1 block every 25 mins (and that is the rate)

[6:59]
Then, even if the Honest miners leave, the SM does not gain more.

csw [7:00 AM]
uploaded this image: image.png
Add Comment

csw [7:01 AM]
The funny thing is that at 50% the SM earns less than a standard attack

At 50%, the SM earns 71.94 Blocks

A standard 50% plus attack gets them 72 (edited)

csw [7:54 AM]
https://en.wikipedia.org/wiki/Base_rate_fallacy
Wikipedia
Base rate fallacy
The base rate fallacy, also called base rate neglect or base rate bias, is a formal fallacy. If presented with related base rate information (i.e. generic, general information) and specific information (information only pertaining to a certain case), the mind tends to ignore the former and focus on the latter.
Base rate neglect is a specific form of the more general extension neglect. (62kB)

[7:57]
https://www.logicallyfallacious.com/tools/lp/Bo/LogicalFallacies/55/Base_Rate_Fallacy
https://www.logicallyfallacious.com
Base Rate Fallacy
Ignoring statistical information in favor of using irrelevant information, that one incorrectly believes to be relevant, to make a judgment. This usually stems from the irrational belief that statistics don’t apply in a situation, for one reason or another when, in fact, they do. (226kB)

[7:57]
Bar-Hillel, M. (1977). The Base-Rate Fallacy in Probability Judgments. Defense Technical Information Center.

csw [8:19 AM]
Oh.... Did we consider what happens to transactions?

csw [8:19 AM]
uploaded this image: image.png
Add Comment

csw [8:21 AM]
The result over time is that you add a queue load factor rho which becomes much larger than 0.5 then the average queue length (Transaction wait time) and response time for messages in the queue get large very quickly.  When rho becomes 1.0, the queue will no longer reach a steady state, but will, in fact, grow in length as long as messages continue to arrive.

There are reasons Bitcoin holds a copy of the forked chain and what occurs when you try and hide this...

[8:24]
AT 33.33% the SM will stop earning profit as the chain degrades exponentially

Do we believe that the other miners will just stop and give up?

They can stop this in seconds.

[8:24]
1.    An introduction to Probability Theory and its Applications, Volume 1, William Feller, Wiley, 1957

phoenix
[8:24 AM]
@antanst

antanst [8:25 AM]
joined selfish-mining by invitation from @phoenix

csw [8:25 AM]
Oh and did you know that there is a very simple code fix for a SM attack (not for revenue, but to try and break the system)?

pesa [8:25 AM]
lol! i can't make out what CSW says, but seems he knows what he talking about ! :grinning:

can't wait for the super secret release

csw [8:26 AM]
It is so very simple.

You are going to kick yourself when it all comes out :slightly_smiling_face:

[8:27]
And I will incorporate a code patch as an IF this ever occurs (as well as how to just use firewalls and a tool to detect it)

[8:29]
The patch is simple.

Include the recorded blocks in the chain tree as a marker of orphans


Use these with the difficulty increase and the orphans that are a part of the SM "attack" are then included meaning the SM attack even if allowed to go on for years can NEVER make a small increase in revenue let alone a profit

[8:30]
@pesa
I am smart enough to have others working with me who can change Craig Speak into Standard English :slightly_smiling_face:

csw [8:34 AM]
uploaded this image: image.png
Add Comment

csw [8:37 AM]
Poisson is simpler to calculate and a fair approximation in certain circumstances...

That is why it is used.

However, when you are making time dependent calculations and testing hypothesis'... you need to be more exact.

Simple approximations do not suffice.

csw [8:37 AM]
uploaded this image: image.png
Add Comment

csw [8:38 AM]
These differences seem minor.
The use of the standard Lambda equation certainly saves time... But, small errors magnify.

csw [8:38 AM]
uploaded this image: image.png
Add Comment

csw [8:39 AM]
These small differences from the Poisson approximation to the Binomial (and later the negative binomial all add up.

[8:39]
Each roll is compounded. It is not one roll and move on the queue, but one after another after another...

[8:40]
Error, multiplied by error, , multiplied by error, , multiplied by error, , multiplied by error, , multiplied by error,

[8:40]
Small at first....

[8:42]
See   “An Introduction to Probability Theory and its Applications”, 2nd Edition, William Feller, Wiley, 1957

[8:43]
*Question*

What does a 2-3% difference do to the model?

tomothy [8:48 AM]
Make the error much much more worse?

csw [8:51 AM]
Yes

[8:51]
Small differences add up when run 1000s of times


tomothy [8:54 AM]
Like a manufacturing imperfection which wears over time due to the defect, exacerbates, and ultimately breaks

csw [8:55 AM]
I am releasing a few papers.

1. A very easy one with a fix and lots of diagrams.
This shows the revenue flaws. Very easy to understand and a simple fix that kills SM even without the error.
Also helps with some Time Warp attacks etc

2. A more complex paper and model
Based on real systems and hosts - incorporating the

3. An SEIR system that defines how the Blocks propagate in detail.

4. The maths. As done from the integrated values (on time censored data)
Sorry, but this one is not easy and will be ignored for the simple stuff by most people

5. The network models. All the data on how the Bitcoin network looks and reacts.


tomothy [8:58 AM]
Anything soon, like a press release, regarding dangers of segwit? Would be timely in light of consensus news. Need a counter narrative starting


cryptorebel [9:18 AM]
segbleed


csw [9:34 AM]
LOL

[9:34]
Soon yes.


pesa [9:37 AM]
segboost

checksum0
[9:38 AM]
asicbleed...

[9:38]
Oh wait...

erik.beijnoff [11:31 AM]
joined selfish-mining by invitation from @bitsko, along with @mengerian

satoshi [12:34 PM]
Looking forward to the release of those papers csw

macsga
[12:36 PM]
just caught up with -most of- the news

[12:36]
they don't know the significance of btc protocol


phoenix
[3:05 PM]
@dimitrist

dimitrist [3:05 PM]
joined selfish-mining by invitation from @phoenix


----- May 23rd -----
satoshi [6:15 AM]
@adamselene

adamselene [6:16 AM]
joined selfish-mining by invitation from @satoshi

adamselene [6:19 AM]
@csw the odds of a selfish miner with 30% hashrate finding 3 blocks before honest miners found 1 block works out to 2.7%?

csw [6:44 AM]
Yes

foorbarbaz [7:00 AM]
left selfish-mining

csw [8:33 AM]
uploaded this image: image.png
Add Comment

csw [8:34 AM]
On pseudo profound bullshit. | Selfish Mining :slightly_smiling_face:

bitalien
[3:40 PM]
joined selfish-mining by invitation from @cryptorebel


----- Yesterday May 25th, 2017 -----
csw [12:41 PM]
Here you go... a Little reading...
https://bitcoil.co.il/Doublespend.pdf

csw [12:41 PM]
uploaded this image: image.png
Add Comment

csw [12:42 PM]
I have been telling you that the maths is an approximation....

csw [1:08 PM]
http://organofcorti.blogspot.co.uk/2015/07/faq-bitcoin-mining-and-luck-statistic.html
organofcorti.blogspot.co.uk
FAQ: Bitcoin mining and the luck statistic
"Bitcoin mining pool, network and exchange analysis" (73kB)

[1:08]
Also got it...

[1:09]
And here is a detailed model

[1:09]
https://arxiv.org/pdf/1702.02867.pdf

zbingledack [2:15 PM]
Oh wow

[2:15]
This seems clearer

csw [2:15 PM]
:slightly_smiling_face:

[2:16]
I like suspense

macsga
[2:16 PM]
@csw did you write that faq (organofcorti)? (edited)

csw [2:16 PM]
No, not I

macsga
[2:16 PM]
he's a very bright man

csw [2:16 PM]
But it is by a maths geek :slightly_smiling_face:

macsga
[2:16 PM]
:slightly_smiling_face:

csw [2:16 PM]
I do recomend the page

macsga
[2:17 PM]
I do too :slightly_smiling_face:

csw [2:17 PM]
So, as I said... wrong math

[2:18]
https://bitcoinwisdom.com/bitcoin/difficulty
bitcoinwisdom.com
Bitcoin Difficulty and Hashrate Chart - BitcoinWisdom
Bitcoin Difficulty hashrate chart and accurate estimated next difficulty.

[2:18]
This is next

[2:19]
https://bitcoinwisdom.com/assets/difficulty/bitcoin-difficulty.png?1495736104 (41kB)

[2:19]
As difficulty is not static, it radically alters the distrbution

macsga
[2:20 PM]
this is correct

[2:20]
more complex calc is needed

[2:20]
may I suggest a FFT?

zbingledack [2:21 PM]
So I gather that the probability of a selfish miner with 40% of the hashpower finding the next block starting from a given time _t_ is indeed 40%

But the probability of the SM finding a certain number of blocks before the HM finds a certain other number of blocks involves binomials

This was not in the SM paper? They used the simplified math?

csw [2:21 PM]
Yes

[2:21]
The simple 10 min exact (Only) approximation

[2:21]
And yes, they did not use the Neg Binomial

zbingledack [2:22 PM]
Oops :)

csw [2:22 PM]
Starting to see what I was trying to get others to see?

zbingledack [2:22 PM]
It at least seems quite plausible that there is something there

csw [2:22 PM]
Sorry I have dragged this on, but I was hoping people would see it without my slapping them in the face with it

zbingledack [2:23 PM]
Funny how you always seem to be keenly aware of what was written in the whitepaper.....


csw [2:23 PM]
Moore's law and increasing competition harm the selfish miner as well

[2:23]
Yes, funny :stuck_out_tongue:

[2:24]
I tried to tell Gun, but he told me how much smarter he was than me... so his loss

[2:26]
If you just model using their model, you fail to test the system

[2:26]
You do not do science, you validate what you are given.

csw [2:29 PM]
uploaded this image: image.png
Add Comment

csw [2:29 PM]
uploaded this image: image.png
Add Comment

csw [2:29 PM]
uploaded this image: image.png
Add Comment

csw [2:30 PM]
Sorry... messy - the edited version is in the office

csw [2:30 PM]
uploaded this image: image.png
Add Comment

csw [2:31 PM]
AND....

This is where I lose people:

csw [2:31 PM]
uploaded this image: image.png
Add Comment

macsga
[2:32 PM]
lol

[2:32]
you actually found one to peer review it ?


csw [2:33 PM]
Yes, it is going into a maths journal (edited)

[2:33]
I could find nobody in the Bitcoin world


csw [2:34 PM]
uploaded this image: image.png
Add Comment

zbingledack [2:36 PM]
Is the "differentiation on a net" stuff necessary to destroy the SM argument, or just icing?

elliotolds [2:36 PM]
What makes this seem plausible to you? You saw how cypherblock's simulations made him think me and peter_r were right, no? Isn't it still the case that no technical person who has looked into this agrees with csw?
zbingledack
It at least seems quite plausible that there is something there
Posted in #selfish-miningYesterday at 2:22 PM

csw [2:36 PM]
uploaded this image: image.png
Add Comment

csw [2:36 PM]
Oh @elliotolds
You still try and protect Gun's model

macsga
[2:37 PM]
My mind is fried; been on a continuous 6h lecture on Bioinformatics & Bioethics

[2:37]
I think I'll pass out

csw [2:37 PM]
CYRIL GRUNSPAN AND RICARDO PEREZ-MARCO ´


[2:37]
They proved it IS a Negative Binomial :wink:

[2:37]
Independantly

[2:38]
They did not then go on to apply it to all the cases I used

zbingledack [2:39 PM]
@elliotolds Seems an easy error to make, especially since the approximation is so close.

csw [2:39 PM]
Oh @elliotolds

You believe that as I cannot simplify this sufficiently for you it makes it wrong..

[2:40]
https://arxiv.org/pdf/1702.02867.pdf

[2:40]
Have a read

elliotolds [2:41 PM]
What specific error do you think is being made, @zbingledack? The only assumption in the paper about mining probabilities is that the probability of finding the next block equals your proportion of the hash power

[2:41]
which you seem to accept

csw [2:41 PM]
And it is wrong

[2:42]
It is an approximate for 10 mins exactly AND only for a following block (edited)

[2:43]
A percent or two here and there, multiplied (edited)

[2:43]
All goes to making it more and more error bound

elliotolds [2:43 PM]
It has nothing to do with the time to find a block. It applies no matter how long it takes. If you buy 10% of lottery tickets randomly and there's one winner, you have a 10% chance of winning. Same concept

csw [2:44 PM]
Then just assume a 1.0 Gamma for the network

[2:44]
You DO understand that it is NOT possible to EVER get a gamma of 1.0 or even be close with pool  mining going on?

[2:44]
@elliotolds
You believe that

[2:45]
Try and selfish mine

elliotolds [2:45 PM]
No matter how long finding the next block takes, if you have x% of hash power you're buying x% of the lottery tickets...

csw [2:45 PM]
And - it is the EASIEST thing to detect and stop

[2:45]
Yes @elliotolds

You know best. Ignore all that nasty maths

[2:46]
http://organofcorti.blogspot.co.uk/2015/07/faq-bitcoin-mining-and-luck-statistic.html
organofcorti.blogspot.co.uk
FAQ: Bitcoin mining and the luck statistic
"Bitcoin mining pool, network and exchange analysis" (73kB)

[2:46]
Ignore these nasty difficult to use Erlang distributions.

[2:46]
Why use the actual thing when you can approximate

cypherblock [2:47 PM]
@csw do you agree that with SM approach, the SM releases blocks more slowly than they would otherwise?

csw [2:47 PM]
They create far more orphans

[2:47]
They cause all miners to lose

[2:48]
And they can NEVER gain more than the amount they could have gained if they did not selfish mine

[2:48]
Here is a web page to try see if you can play with compounding error
http://statpages.info/erpropgt.html

[2:49]
And

[2:49]
http://ipl.physics.harvard.edu/wp-uploads/2013/03/PS3_Error_Propagation_sp13.pdf

[2:50]
And

[2:50]
https://www.lhup.edu/~dsimanek/scenario/errorman/propagat.htm

[2:50]
The best a SM can do id screw all miners (including themselves) up

cypherblock [2:51 PM]
The main flaw with SM is the assumption that high gamma can be achieved, and that other miners can’t do things in defense of it. But I’m not sure @csw you have a case against it if you accept those premises.

csw [2:52 PM]
They can never earn more than the daily revinue, they can earn less but screw up all miners

[2:52]
@cypherblock
Gamma for bitcoin can never be more than 0.0002

[2:52]
And it is costly to make it more than 0.0001

cypherblock [2:52 PM]
my simulations show otherwise. I admit my sims could be wrong. Assume gamma of .5 !!! .5 I tell you :slightly_smiling_face:

csw [2:53 PM]
You are not modeling the network

cypherblock [2:53 PM]
SM miner releases blocks only when honest miners do and at most 1 or 2.

[2:53]
what does this do to difficulty?

[2:53]
assume gamma of .5

csw [2:53 PM]
https://bitcoinwisdom.com/bitcoin/difficulty

[2:54]
https: // bitcoinwisdom.com/bitcoin/difficulty

cypherblock [2:54 PM]
50% of honest miners will receive and mine on top of a selfish block when it is published immediately after an honest block. This is the assumption.

csw [2:54 PM]
The network distance is under 1.32

[2:54]
You CANNOT send to 50% of miners

[2:54]
in a single hop

cypherblock [2:55 PM]
fine but IF you COULD.

csw [2:55 PM]
Then it is NOT Bitcoin

[2:55]
That will be what they make with Seg Wit

[2:55]
In ordere to do that, you need to add hops

[2:56]
The network is far too densely connected.


cypherblock [2:56 PM]
Whatever, but IF gamma is .5 or other suitably high number, then SM strategy might work, don’t you think? Can we not speak in hypotheticals?

[2:56]
if gamma is 0 SM strategy does not work.

csw [2:56 PM]
That will require that you go back to 21 machines and slow networks and even that will only get you a distrance of 3-4

[2:56]
Ok. Are we talking SegWit

[2:57]
That CAN have a high Gamma

cypherblock [2:57 PM]
it is a hypothetical. Assume gamma is .5, what is the result of SM strategy? (edited)

csw [2:57 PM]
SegWit can have a gamma as high as 0.8 (edited)

[2:58]
People are confusing centrality with being centralised

[2:58]
IFF Gamma was to be altered in this way

[2:58]
The SM does not benifit  still (edited)

[2:59]
The network collapses

[2:59]
All lose

[3:00]
Over 50% of all mined Transactions are orphaned

[3:01]
Wallets in SegWit can pretend to be nodes, so in SegWit coin. Gamma can be 0.8 and nobody wins

cypherblock [3:01 PM]
Assume single mining pool is using SM strategy and they are at like 33% hash power. Assume high gamma like .5. Assume other pools don’t detect the SM. This is what I ran my simulations with. It is fine to say those assumptions are completely unrealistic. But given these completely unrealistic assumptions, does SM strategy work? What other unrealistic assumptions must be added?

csw [3:02 PM]
Block Tree

[3:02]
Remember nodes are not JUST a store of the main chain, they also hold memory of orphans and create a tree map

[3:03]
At 33%. What is the most blocks that a miner can create in a day?

[3:03]
Selfish or even Gamma =1

[3:03]
The absolute most blocks?

[3:04]
Averaged :slightly_smiling_face:

[3:04]
And assume 10 mins for now

[3:04]
Average

cypherblock [3:04 PM]
144*.33

[3:04]
if you assume their hashpower is accounted for in difficulty

csw [3:05 PM]
So, if 20% of those blocks are orphaned... Can they solve more?

cypherblock [3:05 PM]
What if their hash power is not adequately accounted for?

csw [3:05 PM]
Explain?

cypherblock [3:05 PM]
Simplistic example.

csw [3:06 PM]
Explain sorry?

cypherblock [3:06 PM]
Imagine miner comes on right now, with 100% of existing hashpower.

[3:06]
Sorry no like 110%

csw [3:06 PM]
A SM or Honest

[3:06]
?

cypherblock [3:07 PM]
Selfish I guess. How many blocks can they mine in a day?

csw [3:08 PM]
So, they add 10% or they are now 110%

cypherblock [3:09 PM]
They have 110% of existing hashpower. Imagine hash power is X, they come online today with X(plus a bit). So total hashpower is now 2X plus a bit.

csw [3:10 PM]
Then they have over 50% of the network

[3:10]
They can do attacks that do not require SM

cypherblock [3:10 PM]
yes, but difficulty is not set for them.

csw [3:10 PM]
And SM earns less than if they just drop all other miners :slightly_smiling_face:

[3:10]
It does not matter

[3:11]
They now have 52.38% of the network

cypherblock [3:11 PM]
My point is that they can mine more blocks than they could if difficulty accounted for them. They could mine 145 blocks in a day, and publish all of those. If difficulty accounted for them, then they could not do this.

csw [3:11 PM]
They can simply mine a block in less time then all others combined

[3:11]
If they added 110%

cypherblock [3:12 PM]
Have to run out for a bit. Think on this. I’m trying to prove the point that if your hash is not accounted for in difficulty then you can mine extra blocks.

csw [3:12 PM]
They have 68.25% of the network

[3:12]
It is still better from a revenue point to just mine and not drop over half the network

cypherblock [3:13 PM]
SM paper works if difficulty adjusts to be lower than what it would be if everyone using honest approach. It is difficulty hiding attack.

csw [3:13 PM]
Assuming a static difficulty

[3:13]
Difficulty grows in a log linear manner

[3:13]
See bitcoinwisdom.com/bitcoin/difficulty

[3:14]
So, in a static world without Moore's l;aw, competition or markets... maybe

[3:14]
But I model this world

zbingledack [3:15 PM]
To be clear, my understanding is that the CSW claim is that SM doesn't even work under those unrealistically-favorable-to-SM conditions.
cypherblock
Assume single mining pool is using SM strategy and they are at like 33% hash power. Assume high gamma like .5. Assume other pools don’t detect the SM. This is what I ran my simulations with. It is fine to say those assumptions are completely unrealistic. But given these completely unrealistic assumptions, does SM strategy work? What other unrealistic assumptions must be added?
Posted in #selfish-miningYesterday at 3:01 PM
(edited)

csw [3:16 PM]
What is the claim to the paper?

zbingledack [3:16 PM]
By "the claim" I meant your claim

[3:17]
(Though see edit)

csw [3:17 PM]
My claim their claim is false.

zbingledack [3:17 PM]
Yeah

csw [3:17 PM]
_"enables the selfish pool to collect higher revenues"_

zbingledack [3:18 PM]
I just think @cypherblock is thinking his question wasn't answered there

csw [3:18 PM]
What is revenue?

[3:20]
_"We show that, above a certain threshold size, the revenue of a selfish pool rises superlinearly with pool size above its revenue with the honest strategy."_

Fair revenue is .33 * 144 = 48 blocks a day

[3:20]
Show me how they earn more than this?

[3:21]
_"We assume that miners are rational; that is, they try to maximize their revenue"_

[3:21]
The loss is in Orphans.

[3:21]
20 or so lost blocks a day.

joeldalais [3:21 PM]
i think he's aiming at - the paper claims that if you withhold a block just before a difficulty adjustment, then difficulty adjustment happens, then you can finish/propagate the block (from before the adjustment changed) faster than the HM
 6 replies Last reply a day ago View thread

csw [3:22 PM]
The assumption here is that difficulty is static

joeldalais [3:22 PM]
yes

csw [3:22 PM]
Try and see what happens when it is log linear....

[3:22]
As Bitcoin is and always has been :slightly_smiling_face:


[3:23]
Now, you also seem to think that the withholding of 4-5 blocks at the difficulty change makes a large difference? Over 2016 blocks and a log linear system

joeldalais [3:24 PM]
they work with the assumption difficulty for their withheld block won't change, that nodes are not 'inoculated' by honest miners (they don't see that the honest miner 'block' is already fast propagating over the network and that the sm block can never ever catch up)

csw [3:24 PM]
Or result in storms of orphans

joeldalais [3:24 PM]
they think that the 'released sm' block (which is late), competes against the 'same level of blocks'

csw [3:25 PM]
And orphans are not revenue inducing. (edited)

csw [3:25 PM]
uploaded this image: bitcoin-difficulty.png
Add Comment

csw [3:26 PM]
Look at the saw tooth effect....

[3:26]
Linear even as an approximation is so far in error

elliotolds [3:27 PM]
@cypherblock you said "if gamma is 0 SM strategy does not work." .. don't your simulations show that even if gamma is 0, SM is profitable if their hash rate is above 33%?

csw [3:28 PM]
Do you understand what profit is @elliotolds ?

[3:28]
If you can NEVER earn more than 48 a day at 33% and you intro orphans

[3:28]
You lose profit

joeldalais [3:29 PM]
they wouldn't be 'profitable' until they have a super majority of hash (which then destroys the system), then it becomes - whats the point of selfish mining? if its for profit, then its more profitable to honest mine

[3:30]
honest mining is always more profitable than selfish mining, short term and long term

[3:30]
occam's razor - if selfish mining actually was legit, then it would have occurred already

csw [3:30 PM]
:slightly_smiling_face:

[3:30]
Max rate - alpha

[3:30]
Never more

[3:31]
Gamma = 1 -> alpha = alpha (edited)

[3:31]
Gamma <1 -> Alpha is < Honest Alpha

[3:32]
The *BEST* you can do is honest


[3:32]
The only *attack* is trying to damage the network and this ALSO has a solution

[3:33]
Scale. Larger blocks, more TXs and scale means lower attack rates

[3:33]
SegWit is higher Gamma.

[3:33]
SegWit INCENTIVIZES  attacks


zbingledack [3:44 PM]
Just to check I have understood the claims:

Even under Segwit we are safe from SM (even if it went undetected), right? It's just that it moves us in the wrong direction by increasing gamma?

[3:44]
Or rather safe from SM being profitable

csw [3:45 PM]
Safe from stollen money

zbingledack [3:45 PM]
So basically SM will never profit, but Segwit does make it easier to use SM to disrupt the network?

csw [3:45 PM]
Not safe from Sybil attacks

[3:45]
Yes

[3:46]
SegWit makes all these Sybil nodes

zbingledack [3:46 PM]
I see

csw [3:46 PM]
The very thing Bitcoin solved, they want to add back

csw [4:27 PM]
Marshall-Olkin exponential distribution

[4:27]
Marshall, A. W. and Olkin, I. (1997). A new method for adding a parameter to a family of distributions with application to the exponential and Weibull families. Biometrika,
84, 641-652. (edited)

[4:28]
A little closer to the real model

cypherblock [4:30 PM]
I am back now. Catching up on above.

cypherblock [5:25 PM]
running a couple more sims to test gamma zero for SM higher than 33%.

[5:29]
to see if @elliotolds is correct about SM approach winning in these scenarios

nicholat [6:22 PM]
joined selfish-mining by invitation from @tomothy. Also, @bergun joined, @hantheme joined.


----- Today May 26th, 2017 -----
csw [3:48 AM]
_“Selfish miners achieve this goal by selectively revealing their mined blocks to invalidate the honest miners’ work. Approximately speaking, the selfish mining pool keeps its mined blocks private, secretly bifurcating the blockchain and creating a private branch.”_

Any miner can do this at any time it is called forking.

The majority group of miners do not do this as their forked chain would have far less value and have a longer transaction rate and not be able to maintain the same level of service as the primary chain. A selfish miner doing the strategy will find very quickly that they are isolated in their own chain. The issue with this is that they will create a separate bitcoin fork and not rejoin with the main chain.

At any time, any miner can choose to fork.

This does not occur as it is not in their rational self-interest. If a pool miner was to do this, they would quickly find that their pool members leave for the main chain reducing their hash rate in the selfish pool further. (edited)

[3:49]
_“ and enables the selfish pool to collect higher revenues by incorporating a higher fraction of its blocks into the blockchain.”_

Again, the authors failed to see the difference between absolute and fractional revenue. They do not understand that a smaller overall revenue comes with the existing costs making a far less profitable strategy.

For instance, if a miner manages to make a little over 10% gross profit, and they have 33% of the hash power, they will expect to earn 48 blocks a day of which just under five blocks a day is attributable to gross profit.

If this 33 percent selfish mines, they will expect to earn a revenue of 42 blocks a day. We have already seen that they are profitable over 43 blocks. Consequently, the selfish mining strategy does not only reduce revenue but it creates a profit loss. The selfish miner moves from a scenario where they are making money to one where they are creating loss. The value of bitcoin is determined at market. As more and more bitcoin are released and the block reward is diminishing, the impact of mining you blocks also diminishes.

[3:50]
_“Its decisions depend only on the relative lengths of the selfish pool’s private branch versus the public branch.”_

The decisions do not depend solely on the selfish mining pools perceived ratio, or mining activity involves profitability calculations. This is a combination of costs and earnt revenue. There are ways to reduce costs. Capital and sunk costs can be attributed over a longer period but operational costs depend on the actions being undertaken. A miner can choose to turn equipment off, lowering the hash rate and their revenue at a particular point in time but at the same time reducing operational costs as they are not paying for the use of the equipment.
This strategy can make sense when there are differentials in operational costs such as off peak versus peak electricity fees.

In the case of the selfish mining strategy, the announced “rational” strategy suggested by the paper’s authors is to maintain costs including operational costs whilst simultaneously lowering revenue. Any first-year business or finance student let alone anyone who has run a company will quickly tell you that this is a recipe for disaster.

Many academics on the other hand who have never worked in the real world see money as a source of endless supply. What they see, is grant funding that comes from a bottomless pool called the government. Because of this lack of real world understanding they fail to understand scarcity. The majority of academics fail to understand the economic effect of revenue versus profitability. To them, sinking more money into something is always good.

csw [4:55 AM]
_“In this section, we argue that the current Bitcoin protocol has no measures to guarantee a low γ. This implies that the threshold may be as low as zero, and a pool of any size can benefit by running Selfish-Mine. We suggest a simple change to the protocol that, if adopted by all non-selfish miners, sets γ to 1/2, and therefore the threshold to 1/4.”_

The authors of the selfish mining paper have completely ignored testing anything to do with the network or its topography. They have made assumptions and have failed to understand anything to do with how the machines interconnect.

The gamma of the real network is set to be maximised at a rate of 0.002 (in reality as as Bitcoin is constructed) and no rational decision would be likely to ever create a scenario where more than a gamma of 0.0001 can exist. The so-called proposed change would actually increase the likelihood of the network being attacked. There assumption that their change decreases gamma is false. The solution proposed by the authors of the selfish mining paper creates the scenario that they are seeking to model.

The ONLY way a Gamma can be at 0.5 is to implement the absolutely asinine _solution_ proposed in that paper.

[4:58]
*Detecting selfish mining*

In contradiction to the comments that it is difficult to detect the selfish mining, I will counter this with the notion that the authors of paper are dangerously deluded by their own model.

Any selfish mining attack is *trivial* to detect (visually by a 3 year old with no training) and block (resulting ion a complete revenue loss to the SM. The current model does not drop orphan blocks but rather holds a block tree in memory. In failing to analyse the protocol or the software the authors have ignored the actual system for their mythical idealised system that they hold in their head that is never existed.

The authors of the selfish mining paper choose a version of reality dictated by Aristotle. They seek to model a perfect scenario in their head without ever checking reality. They see a perfect version of their ideal of bitcoin and not the code or the protocol as it actually exists. In this, they model a circle and talk about the perfection of the circle and how it could be better. Unfortunately the closest analogy would be that the item there modelling is a cube that they are talking about as a two-dimensional circle. The authors are simply wrong and they are completely ignorant.

The authors of the selfish mining paper are *dangerously deluded*, *horrendously unscientific* and their processes and failed to comprehend the basics of developing a scientific experiment.

[4:58]
As for detecting hidden patterns:
https://uk.sans.org/reading-room/whitepapers/stenganography/detecting-hydan-statistical-methods-classifying-hydan-based-stegonagraphy-execut-32839

[4:59]
I used to write code to detect Steganographic information.
https://en.wikipedia.org/wiki/Steganography
Wikipedia
Steganography
Steganography (/ˌstɛɡ.əˈnɒɡ.rə.fi/, STEG-ə-NOG-rəfee) is the practice of concealing a file, message, image, or video within another file, message, image, or video. The word steganography combines the Greek words steganos (στεγανός), meaning "covered, concealed, or protected", and graphein (γράφειν) meaning "writing".
The first recorded use of the term was in 1499 by Johannes Trithemius in his Steganographia, a treatise on cryptography and steganography, disguised as a book on magic. Generally, the Show more… (2MB)


[5:00]
I was the one who ruined Trucrypt.


[5:00]
http://www.ibtimes.co.uk/truecrypt-goes-way-lavabit-developers-shut-it-down-without-warning-1450467
International Business Times UK
TrueCrypt Goes the Way of Lavabit as Developers Shut it Down Without Warning
The encryption service used by Edward Snowden and other privacy-concerned citizens has shut down without warning (88kB)
May 29th, 2014 at 10:37 AM

[5:02]
Some of the fragments of my papers and reports (though this was a draft - no idea where this was leaked...) on this topic are floating about
https://www.homeworkmarket.com/sites/default/files/q4/03/09/samplereport1.pdf

[5:02]
I developed the statistical tools that broke the so called hidden partitions in TrueCrypt.

[5:03]
So, I will say with certainty (as TrueCrypt and Hydan are hard, Selfish mining is trivial.

[5:03]
Trivial for a person who has never used a computer level trivial.

cypherblock [6:51 AM]
@csw if you are writing a paper to refute SM strategy, please take into account difficulty adjustments caused by SM strategy.

csw [6:52 AM]
I am

[6:52]
I did note the saw tooth structure of mining difficulty remember

[6:52]
It is Log Linear with periodic adjustments...

[6:53]
I believe that I noted that, and this makes the assumptions incorrect in the model used

cypherblock [6:58 AM]
well I guess my point is, if everyone’s hash power is fixed, and a 33% pool starts mining selfishly, will the difficulty go up, go down or stay the same? To me it looks like difficulty adjust downward because rate of blocks published is lower.

csw [6:59 AM]
Poisson is not an approximation in this instance

[6:59]
Not even close

[7:00]
2016 blocks - the time is longer

[7:00]
The number of blocks is the same :slightly_smiling_face:

cypherblock [7:01 AM]
yes, takes longer to get to 2016 blocks, so won’t difficulty adjust downward to make this happen sooner?

csw [7:01 AM]
At 33% the Selfish miner loses more to orphans then the honest miner

[7:02]
It is a Black Swan form of error.

The distribution is asymmetric. It is not distributed evenly around the means

[7:05]
@cypherblock
] Marshall, A. W. and Olkin, I. (1997). A new method for adding a parameter to a family of distributions with application to the exponential and Weibull families. Biometrika, 84, 641-652.

[7:06]
This is still an approximation, but it is far closer

[7:06]
http://home.iitk.ac.in/~kundu/moged-rev-3.pdf

cypherblock [7:07 AM]
While I appreciate all the references, the question remains, won’t difficulty adjust downwards in the scenario I just described?

csw [7:07 AM]
It is probabilistic

[7:07]
So, are we talking once, on average, over several changes?

cypherblock [7:07 AM]
over several changes let’s say

[7:08]
although I’m pretty sure even the first change has high probability of being affected

csw [7:08 AM]
The time will get shorter, so the difficulty is lowered across several "standard" changes - if these can be said to exist

[7:08]
Here is the thing with the Negative Binomial model

[7:09]
When the time to discovery is shorter, that is lambda is lower, the distribution favors the honest miner

[7:09]
When it is greater, it favors the attacker

macsga
[7:09 AM]
good morning

[7:09]
https://i.imgflip.com/1pr66c.jpg (14kB)

csw [7:11 AM]
LOL

[7:11]
@macsga Good one

macsga
[7:11 AM]
:wink:

csw [7:13 AM]
uploaded this image: image.png
Add Comment

csw [7:13 AM]
https://arxiv.org/pdf/1702.02867.pdf

[7:13]
That is not my paper, that is what using the Negative Binomial does

[7:15]
Here is something that you may have not gotten with the Negative Binomial, as the  difficulty decreases, the time to find a block is lower. For all parties.

When this occurs, it is the Honest miner who has the greater advantage. It is not symmetric.

cypherblock [7:16 AM]
Anyway, so difficulty adjusts downward. This effects # of reward blocks the SM miner can produce in a day (yes effects everyone else too).


csw [7:16 AM]
uploaded this image: image.png
Add Comment

csw [7:16 AM]
Yes. And to point you to a paper and a quote:

"Therefore the correct results for bitcoin security are worse than those given in
[7]. The explanation is that Nakamoto’s result is correct only if the mining time by
the honest miners is exactly the expected time. *Longer than average times help the attackers*"

[7:17]
As the time decreases, the selfish attacker loses...

[7:20]
What happens - 33%

1. Selfish Miner loses orphans - at 33% just slightly more than the HM
2. Difficulty adjusts - this makes the discovery faster
3. The HM mines far more than the SM who loses MANY blocks at an increased Orphaning
4. Difficulty adjusts
5. Selfish Miner loses orphans - at 33% just slightly more than the HM

...

looping

[7:22]
At 40%

1. Selfish Miner loses orphans - at 40% just slightly *less* than the HM
 - lowewr revenue, but a slight ration gain
2. Difficulty adjusts - this makes the discovery faster
3. The HM mines far more than the SM who loses MANY blocks at an increased Orphaning
 - The SM loses more than the HM
4. Difficulty adjusts
5. Selfish Miner loses orphans - at 40% just slightly more than the HM...looping


...

Loops

cypherblock [7:22 AM]
difficulty adjust will stabalize. Compare SM blocks per day after difficulty adjsuts and stabalizes, vs SM pool mining honestly

csw [7:22 AM]
At 44% - the system becomes chaotic

[7:23]
No, it actually becomes cyclic at 33% and chaotic above 44

[7:26]
And @cypherblock You are modeling with Poisson assumptions

[7:26]
These are not valid assumptions. (edited)

csw [7:44 AM]
uploaded this image: image.png
Add Comment

csw [7:44 AM]
uploaded this image: image.png
Add Comment

csw [7:45 AM]
These are the long term effects of difficulty and selfish mining.

[7:45]
The results are chaotic - as in they express fractal dimensionality

[7:45]
Over 44%, the chain cannot be predicted.

csw [7:47 AM]
uploaded this image: image.png
Add Comment


csw [7:47 AM]
This is what I mean by periodic.

[7:48]
Or the basic (hand drawn) version

csw [7:48 AM]
uploaded this image: image.png
Add Comment

nicholat [7:49 AM]
please explain to me the subject of this debate and channel like I'm 5 years old? :stuck_out_tongue:

csw [7:49 AM]
Selfish mining is not profitable

csw [7:49 AM]
uploaded this image: image.png
Add Comment

cypherblock [7:50 AM]
@nicholat discussing whether selfish mining approach (described in paper by Guy Sirer and Eyal) works or not

csw [7:51 AM]
At 44% you simply cannot say what will occur. Small differences in the inital condition or changes in hash rate or discovery lead to Bifurcation

[7:51]
A  maximum Lyapunov exponent is created.

I cannot explain this to anyone without using maths

nicholat [7:52 AM]
well, the gist of it?

[7:52]
I mean, try to explain it as you would to your mother

csw [7:52 AM]
uploaded this image: image.png
Add Comment

csw [7:53 AM]
That is at 45%

[7:53]
Completely unable to be predicted.

cypherblock [7:54 AM]
@nicholat : selfish mining strategy means instead of publishing blocks as soon as you find them, you build a private chain of blocks and only publish your blocks to the bitcoin network when the remaining honest miners find a block. Sometimes you publish 2 blocks as needed to confirm your rewards and beat the others.

nicholat [7:54 AM]
so basically, "selfishly mining" can cause problems with less than 51% of the network hash rate?

[7:54]
right, I've heard about that

tomothy [7:55 AM]
The paper alleges miners can cheat and if they cheat they have more revenue. It assumes revenue  is what drives miners not profit. Revenue doesn't help  much if you aren't profitable. Their study is based on assumptions. (edited)

tomothy [7:55 AM]
If the assumptions are wrong, their study is wrong. (edited)
1 reply Today at 6:11 PM View thread

nicholat [7:56 AM]
right, and that's what csw is trying to find out about

tomothy [7:57 AM]
More or less yes. He shows that they used simple math.

csw [7:57 AM]
I have papers in review already

tomothy [7:57 AM]
For a complex problem.

csw [7:57 AM]
I am trying to teach others :slightly_smiling_face:
1 reply Today at 6:13 PM View thread

csw [7:58 AM]
uploaded this image: image.png
Add Comment

tomothy [7:58 AM]
And ultimately, you can't cheat maths.


csw [7:58 AM]
That is what happens at 45%

[7:58]
as a Selfish miner

nicholat [7:58 AM]
right; I mean, I have a few years of university calculus and linear algebra behind me, I understand some of the terminology you're using in the paper

[7:58]
but definitely not all of it

mai8304 [7:58 AM]
joined selfish-mining by invitation from @tomothy, along with @demoforyelin

csw [7:58 AM]
I needed to have a pure maths journal for publication.

cypherblock [7:59 AM]
@csw what is that chart? What is it showing. Looks impressive, but useless without explanation

csw [7:59 AM]
I was unable to find peers to understand the maths in the Bitcoin community. We all have different skills
1 reply Today at 6:00 PM View thread

csw [7:59 AM]
That chart is the blocks over time won by the selfish miner when the difficulty is accounted for

[8:00]
It leads to Bifurcations and Chaos and is a Periodic Model

[8:01]
That image is what happens to the chain at 45% if the honest miners all allow (and they can stop this at will) the chain to have selfish mining and they have only 55% as honest miners

[8:02]
They *may* get a lot of control. They *may* get nothing

More than not, they lose

[8:02]
I do not believe that a rational miner will engage in a strategy that will cost them money. The paper stated they would

new messages
weekend_engineer [5:04 PM]
joined selfish-mining by invitation from @tomothy, along with @daganb

newliberty
[5:51 PM]
uploaded this image: Pasted image at 2017-05-26, 2:50 PM
Add Comment

newliberty
[5:57 PM]
It applies to a lot more than selfish mining, as do the other elements used in the proofs, (or more accurately, the disproof of the conjecture).  The SM things here are basics.  An application of principles of maths.  If you get all of it, and can show the work, it may qualify you for going where the angles fear to walk. (edited)

newliberty
[6:02 PM]
uploaded this image: Pasted image at 2017-05-26, 3:02 PM
Add Comment

nicholat [6:17 PM]
looking at your interviews from 2014 btw @csw, what's your current take on PSOs?

[6:17]
I recall an optimization problem I never found a good method for, which I now realize I could possibly have used it for (edited)

btcalbin [6:20 PM]
Kind of non-technical intuition, but I have the feeling that the general class of block-withholding attacks and Maxwell's conception that without a block reward, the chain will not move forward (because of miners constantly reorging at the same height to take each others fees) have something very significant in common. They both seem to be predicated on the idea that a malicious miner can just keep attacking forever without actually taking the money, as though there is no time value of money.