Why Distributed Reputation is Immune to Sybil Attack.

in reputation •  8 months ago 


I've been trying to explain my idea for a reputation system on the Steem blockchain but everyone asks how it won't be overtaken by bots.

Basically, every account gets a say of who they trust. If someone you trust then trusts someone else within the system, that information is broadcasted to the network and you'll be able to use that data to expand your trust network.

network decentralized trust reputation.jpg

So, take the situation where one person creates a thousand bots that trust himself. This doesn't get the botter anywhere because no one trusts the bots. He's back at square one trying to get the accounts that matter to trust him. The intention of the entire system is to create incentives where your reputation is worth more than what you can gain by cheating the system. (Just like proof-of-work coins.)

In order for trust and reputation to be earned within this distributed system, someone in your network has to trust the person in question. However, how much you trust someone in your network depends on a lot of variables. For example, you might be very willing to trust the judgement of say 10 or 20 accounts on Steem, but are you willing to trust the accounts who those people trust? Maybe. Maybe not. It all depends on how the individual's user configuration is set up to handle diminishing returns down the line. Each leap makes a big difference.

Network effect

I haven't even started trying to program this reputation system because in order to have value a lot of people need to be using it all at the same time. Therefore, I'll have to make something useful to bootstrap the project; something that gives the project value before the network effect kicks in.

As of now I'm looking at custom (distributed) trending tabs and peer to peer vote buying/selling. These two issues are the real crux of the Steem platform. Steemit's poor trending tab creates the demand to buy votes for more than they are worth for very little payoff; leeching the system. I don't see the other frontends stepping up to this challenge either to make better trending tabs.

As I side note, I think it's obvious that ad revenue should also be shared with content creators. The platform becomes outward facing when content creators have a financial incentive to bring in communities from outside the cryptosphere.


By creating demand to buy votes and then monopolizing the market with centralized bid-bot services this platform is being a bit taxed, and everyone complains about it. I would say it's not a big of a deal as people make out, but it obviously is a problem.

The solution isn't to double down on curation, create a new downvote pool out of thin air, and manipulate the reward curve to punish low vote posts/comments. The solution is and always has been to decentralize the process of vote buying/selling and create custom trending tabs.

It's actually a little bit embarrassing that Steemit Inc hasn't allowed people to buy/sell upvotes directly on the internal market. Clearly, there is great demand for that service. Ironically, the demand is created by Steemit itself, but they somehow refuse to create the market that they themselves produced the demand for.

Instead, they are allowing the big players of Steem to run exploitative bots that siphon inflation in a centralized manner. Why hasn't this obviously trivial process been decentralized yet? It's the same kind of liquidity issue as buying and selling crypto directly.

So, I've said this a few times... I'd like to create a decentralized way to buy/sell votes. One with a free market. One where anyone on Steem can theoretically make a profit instead of just the bid bot owners.


In relation to the reputation system I want to make: Say level 0 is the baseline. Everyone starts at level 0 by default. Everyone who you follow gets promoted to level 1 by default. Now, lets say you want to make a contract that allows level 2 users on your list to buy your vote at a discount. Say you demote someone to level -1 for whatever reason. People in this category would have to pay you more to buy your vote.

This is just one application of the distributed rep system. The reputation can also be used to create the custom trending tabs. Currently, trending tabs only use two variables: payout amount and time. The higher the payout and lower the time the more "points" the post has and the higher it appears on the trending tab.

Well, what if you start using more information to filter your trending tabs? What if people start sharing a blacklist of accounts that allow their vote to be sold to posts that we don't want to see anymore? What if accounts on our whitelist start counting for double or triple their original value? These are the kinds of things we should be tinkering with on the platform, and they are completely permissionless.


Reputation as a product

What happens when a vendor selling a product pays confirmed buyers to give them a five star review? The users give the five star review. Why wouldn't they? With decentralized reputation you can hold accounts responsible for the reviews they make.

You told me XYZ was a good product. It was garbage.

Now reviewers have a real reason to not sell out.
They're being held accountable because if they give a bad review their own reputation will be affected.
They actually have something to lose.


And that's what reputation is supposed to be all about. Finding trustworthy people and holding bad actors accountable. I search far and wide on this platform, and I see no one working toward these goals.

That's why I've taken such a big risk on PALNet. If I can bring some value to this network it won't matter if the products that I create generate direct income or not. If we all do our jobs properly, everyone should benefit the entire ecosystem. Synergy and network effect reign supreme in the cryptosphere.

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!
Sort Order:  

This is basically the same idea as Larry Page's PageRank for websites.

In his model, each link from one HTML document to another was like a vote of confidence.

Over the years that has been refined to include how relevant one thing is to another, the authority each party has in general as well as on the specific topic.

So all I need to do is apply my SEO skills to steem?!? Done!!!

The analogy holds, yes.

In SEO, you are trying to give the SE what it wants. You want to be a trusted source that is promoted by the search engine. The same holds true under this kind of reputation system.

Your on-page SEO is sculpted to showcase domain expertise and a lot of it. Similarly, in this rep system, you would want to publish content that makes others view you as a trusted source in whatever area it is you are publishing in. You also don't want to spend that trust by linking to a bad source.

Your off-page SEO is an even stronger analogy. The Sybil attack that @edicted is talking about is, in SEO-terms, a low-quality PBN. Just spinning up a thousand accounts or a thousand blogs and linking back to yourself does not work because those feeder accounts don't have any authority on their own.

The side effect of this is that high-quality, relevant PBNs are still effective and worth a lot of money. In this new rep system, a high-trust account could be sold along with its authority to a bad actor. Hopefully, if that happens in a way that produces a really negative result people could respond and remove their support.

  ·  8 months ago (edited)Reveal Comment

It has been determined that you are trash, therefore, you have received a negative vote.

PLEASE NOTE: If you engage with the trash above you also risk receiving a negative vote on your comment.

These are worthwhile ideas. The reputation system based on upvote history is obviously worthless. User Authority is slightly better but the idea is sort of half-baked and and the connection between engagement and rewards is so unclear as to not drive much in the way of increased engagement.

A trust/endorsement system is a very interesting idea because it relies on human interactions and judgement. A trust-based system like that could serve as the basis of an account based curation system for one thing.

This reputation system you propose is very interesting.

In regards to trending among those you "trust", I and others have suggested similar going back over a year ago. The basic idea would be things trending among your friends. Basically a simpler idea along the same lines of what you said.

I think if Steemit had actually implemented some of the ideas like this and others quite a while ago, we'd be in a very different situation now, because you could see what was trending among your friends, rather than just what some random whale or bot upvoted.

I'd also like to see more curation tools built into the interfaces. Curation should mean more than just upvoting, but actually finding awesome posts that other people will love.

  ·  8 months ago (edited)

I've always liked the decentralized reputation system from the fist time you wrote about it. The problem is that Vote selling will not be something that a large part of authors will take part in, not only that, but it undermines the Rewards, and they no longer are rewards, it also will undoubtedly drive most content creators away from here, with them, most content consumer's, all because of the consequences of costly exposure, which will be costly not necessarily because of the demand/supply but because it will simply cost, it will not be free. So then who will populate these custom trending tabs, and who will consume them, especially when there's free alternatives.

Ideally custom content discovery would completely undermine the need to buy/sell votes entirely.

  ·  8 months ago (edited)

I'm confused. Not trying to pick apart what you just said but I'd like to tie back in to what I was saying:

The need to buy votes is undermined by the free votes, but the want for exposure by buying votes will attract only certain people, and hardly any content creators (starving artists will starve) but more likely advertising/marketing types (do you know about Jerry?), and of course, the fucking dread of the crypto world, the TA's, which I forsee as driving most of the social life straight out of the "market", and as I understand, the trending tab/custom content discovery Is the market, so unless I'm confused about the last part I don't understand how the product of vote buying will stop vote buying?

I figure if we find new heuristics to discover content that don't rely on post payout then there will be little incentive to buy votes in the first place because curation is happening without it.

OK now I understand it, the only question now is how do we avoid the circles being closed and encourage content discovery outside them, from new users not yet vetted? I imagine it would have to be through the "created" tab which poses the issue of not having enough attention/time to scroll through it even, let alone read the content, if and when the traffic explodes. I understand that certain known authors will be vetted by one circle or another as soon as they enter the scene, but what about a nobody?

Posted using Partiko Android

Closed circles are a part of tribalism and hard to avoid. I would hope it would be financially viable to expand one's trust network and continue to seek out new users because the network itself gains value from such action, personally benefiting the stake holder in question.

I also think curation should revolve much more around resteems, as reblogging is practically the definition of curation to begin with. I think ad revenue needs to be split with content creators to face the frontend outward.

I think a new version of the created tab could be modified to filter it using far more metrics than just a timestamp. There is a vast wealth of information just sitting on the blockchain that remains completely underutilized.


  ·  8 months ago Reveal Comment
  ·  8 months ago (edited)

Ok, the bots are all connected.


Nothing happened.
No important accounts trust any of the bots.
No one cares about this botnet.
This botnet still has zero reputation within the distributed system.
It is contained... quarantined even.

  ·  8 months ago Reveal Comment

It has been determined that you are trash, therefore, you have received a negative vote.

PLEASE NOTE: If you engage with the trash above you also risk receiving a negative vote on your comment.

Which socks do you think I use? I mainly have this account and one other for games only, and am starting to diversify my funds to yet another so all isn't in one. So just three and I only interact with this one.

  ·  8 months ago (edited)Reveal Comment

It has been determined that you are trash, therefore, you have received a negative vote.

PLEASE NOTE: If you engage with the trash above you also risk receiving a negative vote on your comment.

Sure no worries, can you go back to the @kawaiicrush post, I have left some detailed replies there, the 200 Steem received from the steemit account was a bounty from Ned for the first x amount of people to review the Steem 101 Ebook on amazon, quite a few other people received it as well.

Once that is out of the picture there is nothing left to tie me to Steemit, I certainly have no professional affiliation with them.

It has been determined that you are trash, therefore, you have received a negative vote.

PLEASE NOTE: If you engage with the trash above you also risk receiving a negative vote on your comment.

It make sense to me in the same way google uses link attribution as a major ranking factor!

It also allows you to create thousands of patterns so scammers can get picked up and punished easily and keeps people honest!

Sure you can try to talk to accounts and built a circle jerk which people might do but it would be more effort than just being a good steemian

Posted using Partiko iOS