Implications of a 76% attack


#1

I have some confusion regarding the security of the network and would really appreciate someone more knowledgable than myself having a stab at a few questions which entered my mind:

What is the likelihood of a 76% attack at any point?

What are the practical implications of such an attack?

Can information held on the network be deleted?

Can information held on the network be unencrypted?

Would ’the attacker’ be able to re-disribute or somehow disrupt safecoins held on accounts?


#2

Quick answers (sorry, running like mad spent all day talking with devs so need to work).

1: It will be harder as time goes on for many reasons, but the age of vaults does make a difference. Please see attacks part of systemdocs

2: It depends on severity, most likely attack is denial of service or vandalism

3: If the attack is large enough then maybe (we have run a network of over 23,000,000 attacks against a network of 10,000 and could not do this). I suggest an attack of this size would be undefined behaviour type of thing.

4: Absolutely not :slight_smile: this part is cool the network has no ability to decrypt data.

5: This is like 3 but worse for the attacker, if this can happen once then safecoin value would be zero perhaps. There is a huge amount to say about this though. I will one day when I can breath, but basically an attack of that size would be like a 51% bitcoin attack I suppose, but … unlike bitcoin it may be limited to certain coins. I will go into as much as I can though as we go through testnets

Sorry for brief answers but the questions are small and answers are actually huge (like why is the sun there kind of thing)


#3

Thanks for the answers David!! I think you and the team are doing something really important and significant so keep up the amazing work :smile:

It is my understanding if there was a problem (namely a 51% attack) in bitcoin then the community could potentially form a consensus opinion regarding a moment in time when the block chain was disrupted, and ‘re-boot’ the chain from that moment whilst removing the ‘bad actors’. My understanding is that due to the architecture of safecoin that no such option is available. In other words the significance of an attack could be far greater.

Whilst the fiat value might indeed become pretty much zero after such an attack, as long as there was some way of re-assigning safecoin to the correct accounts (prior to the disruption) then the network could pick itself up, dust itself off and move forward. If however there was no such ‘backup’ then safecoin would essentially remain at zero value and a brand new network would be required instead of an improvement/tweak to the original. For example no-one would want to be developing apps for a network where someone had stolen and controlled 90% of the ‘wealth’. Could there potentially be an encrypted blockchain (or something similar) for this purpose or am I talking nonsense?

If anyone has any thoughts about this I’d love to hear them. There’s so much to wrap your head around with all this stuff!!! Not surprised it’s been growing and evolving for 8 years:)


#4

Rolling back the blockchain not only removes bad transactions, but also good ones too. If the good ones were taken in full faith and goods had been exchanged, rolling back is going to harm them.

So yeah, it is an option with Bitcoin, but not a desirable one. Just thought I would mention this for comparison reasons.


#5

Also, if someone actually does possess 51% of the hash power they can do the same thing again. There is no obvious way of shutting certain miners out.


#6

This is a quote I found from Andreas Antonopoulos:

“I do not worry about a 51% attack on bitcoin. It is neither likely, nor effective”

Given that some people have their entire net worth on the block chain I’m sure this is pretty comforting. At least they’d only lose the fiat value of their bitcoin and not the bitcoin itself (which one assumes would potentially re-gain some sort of value).

It is not clear (at least to me) what the severity of a compromise in maidsafe could (in the worst case scenario) be for safecoin (or indeed what form it could take). Without a ‘rewind button’ the safecoin economy could potentially disappear entirely in the event of a successful attack, dragging the entire network down with it. It seems sensible to have a ‘nuclear bunker’ for safecoin just in case, does this currently exist? Is it indeed feasible? Am I babbling nonsense? :slight_smile:


#7

Please keep in mind the biggest difference betwen safecoin and bitcoin :
unlike bitcoin having a centralized info bookeeper, safecoin doesn’t have such thing.
Every safecoin is atomic and can be considered as a kind of blockchain itself.

Even someone managed to lauch a 76% attack, the only achievement will only be assign couple of safecoin to himself (it is under consideration to handle this edge case).
There won’t be any damange to any existing safecoin, whilst no damage to any existing account.


#8

Could someone elaborate on exactly where the number 76 comes from?

Also, when discussing numbers like this everyone should be aware that in the Bitcoin case we talk about hash power while in the SAFE case we talk about number of nodes so they can’t really be compared.


#9

[quote=“Luke, post:6, topic:910”][quote]
I do not worry about a 51% attack on bitcoin. It is neither likely, nor effective
[/quote]
Given that some people have their entire net worth on the block chain I’m sure this is pretty comforting
[/quote]

I agree with Andreas, it isn’t likely. That’s just one person’s opinion though (ok, two), others are more worried. It would take a lot of money and someone who simply wants to destroy/discredit bitcoin, but it is doable.


#10

Sorry, that was me :slight_smile: I saw someone talking about the ramifications of one ‘bad actor’ controlling more than 75% of nodes they could take some degree of control of the network. I figured it would make sense to coin the term ‘76% attack’ in line with the naming convention assigned to the bitcoin threat (albeit I understand they are very different forms of attack).

So does that mean it would be literally impossible for an attacker to get at/redistribute the safecoins associated to any account? Also, from where would the coins which the attacker assigned to himself come from (if not the accounts/nodes where funds are being held)?

I’m sorry, I don’t understand what this means. Are you able to elaborate at all?


#11

This is a very important distinction, thanks @qi_ma for pointing it out. To expand on what this means:

  • bitcoin 50% attack: allows the attacker to control confirmations, insert fake transactions, steal any amount of bitcoin they choose from anywhere on the blockchain. Of course, this would be disasterous for bitcoin which makes the attack self defeating, but of course still a concern.

  • safecoin 75% attack: would give the attacker control of enough adjacent nodes to reassign a small number of bitcoin (@qi_ma mentions a couple) controlled by those nodes - it doesn’t affect all safecoin, or enable them to decrypt information on those nodes. In this case the high cost and small benefit make the attack infeasible, but of course still a concern.

@David: the figure of 75% (or more) for the safecoin attack comes from analysis done by MaidSafe that @dirvine has referred to, but I don’t know if this is in the documentation or papers yet. I could probably find some quotes, but for now I think we can just take it as David says! :slight_smile:


#12

Yes the 75 refers to the point where an attacker could vandalise or perhaps get access to all copies of a chunk. The safecon attack is nowhere near this figure though, it is likely multiples of the network population (several factors perhaps). So 75% is roughly enough to gain control over the holding of data (pivot point) as I said elsewhere though this is 75% of the rank.

A good point soon the SAFE network will have a 75 attack listed (people need figures like mad) , when in fact 75% is the minimum vandalism / DOS number, its very much conjecture though. It will have to be 75% of the nodes run since day 1 and behaving then all of a sudden changing. It will only vandalise and not do anything to safecoin (bar potentially blocking access to check a coin, but not alter it), that’s a measure larger than 100% of the network population.


#13

I don’t understand this, can you explain what this means. Multiples of the network population I can’t make sense - it leads me to think the attack is impossible! How can anyone own more than 100% of the network? Obviously I have the wrong end of a very big stick here… confused!


#14

No worries

Network size == 100
attack size == 1000 (approx) for a linear addressed network consensus chain.

For our purposes it will be much larger than this. You cannot join networks like that so the 1000 attack nodes have to join over time and gain rank to overpower the 100 nodes and hope the 100 nodes do not increase in number etc.


#15

Ok, so you are using different terminology for the two attacks I think. Just to be sure I’ve got this.

When you talk of a 75% attack this means a network size N contains 0.75 x N attacker nodes plus 0.25 x N regular nodes

A “ten times network size” attack contains N / 11 regular nodes and ( 10 x N ) / 11 attacking nodes and could be called a 10 / 11 or 90.9% attack.

My head hurts :-/


#16

Yes that’s it Mark :smile: Its hard to get right. In Bitcoin you see if you have 100pH then you need to add in another 100 to get to 50%, but as its only hashing power (no notion of rank or how long you have been hashing, just pure add in power) then its significantly easier. The rank differentiates us an awful lot, so the pure numbers make it very hard to equate.


#17

In the 50% bitcoin attack you can’t steal bitcoins from other wallets because you need the private key to do that. You can double spending or prevent people to send bitcoin but not steal any amount of bitcoin.

For the 75% maidsafe attack, as I see it, the problem for the attacker is that not only need a minimum of 3x nodes that the rest of all network size but gain rank. So the attacker need first to contribute in the success of the safe net. This success will made that new nodes were create who increase the need for the attacker. Is a vicious circle for the attacker who have more possibility that, in the end, will benefit the safe net instead of control it.


#18

@digipl yes indeed, thanks for the correction and clarifications.


#19

Wow, that’s genius.

So, the only person / group who could end up harming / hacking the SAFEnet would be the group that has the most invested in its SUCCESS…

Pure genius!!


#20

Just to be safe could we not put an option in the client that allowed users to skip layers of closest nodes while looking for nodes to connect to? If this whole attack hinges on an attacker controlling the nodes closest to a user, their inner layer so to speak, would it not solve the problem just to be able to skip that layer? I’m still having trouble wrapping my brain around how an attacker could control 75% of the ENTIRE network that spanned the globe.