[Workstream] Fraud Detection and Defense Working Group Assemble!

Hey everyone, here is Robert. I have worked on collusion analysis in Gitcoin Grant Round 3 before. Happy to contribute to the workstream :slight_smile: Will try to join the next meeting

3 Likes

There are some good discussions happening in the sybil-defense and grant review channels. Slowly slowly slowly things are clearing up. 4 sub-work groups are being defined. Whether you are a data scientist or a cop Iā€™m sure your experience can be valuable. I look forward to seeing everybody at the next meeting.

1 Like

I have a few suggestions on this topic:

  1. Adjust the meeting time, whether it can be postponed for 2-3 hours, because I slept soundly during the meeting time;
  2. I have seen someone at the 9th and 10th sessions I donated more than 100 accounts, I think this conference will become more important;
  3. I suggest that TRUST BONUS certification should be re-certified every session, it is best to re-examine with BrightID, etc.;
  4. I think the proportion of donated equity needs to be adjusted , If the distribution of bonus needs to evaluate the accountā€™s contribution based on community activity, this will require a function.

If you agree or disagree with my ideas, please express your opinion. I very much hope to hear everyoneā€™s voice.

3 Likes

Hi @robertyan thanks for writing! The next workstream meeting is scheduled for next Tuesday, July 6 at 4pm EST / 10pm CET. I have asked that you are added to the calendar invite.

We are holding a closed Evaluation Panel meeting (due to data privacy issues) tomorrow for the working group who is judging sybil flags for this round in the place of the regularly scheduled meeting.

Thank you again and look out for updates, posted on this thread when there is news to announce :pray:

2 Likes

We have put up a grant to fund the work of this group, would super appreciate your support and sharing in your networks :pray: Thank you so much!

2 Likes

I got the First Contribution! :slight_smile:

I just shared this post on Linkedin here:

2 Likes

Iā€™ll like to join the ML/DS work-stream based on experiences in Data mining and ML Dev and applications in different fields. Though this background is an unfamiliar territory for me, Iā€™m sure Iā€™d impress upon the team toward getting results

1 Like

Since I canā€™t participate, thank you very much for this video

2 Likes

How do I get involved, Is the worksheet still open for new entries?

1 Like

We may not be able to totally eliminate these unwholesome practices, but we can make it too expensive so that it becomes less unprofitable. the use of BrightID and other such verification will go a long way to solve the problem.

3 Likes

Check this out @Luvlynj
Would really be nice to contribute to this brilliant idea

1 Like

Hi everyone! @andajun @Imuetinyan @Luvlynj @Cryptosense if you are interested to participate, please input your info in the spreadsheet above and you will be added to future calendar invites.

We are on a few weeks break while we integrate and document learnings from Round 10.

BlockScience just released this article outlining the teamā€™s work and the work of the Anti-Fraud working group in Round 10:

@DisruptionJoe will be providing further updates about working group meetings and plans going forward in the next weeks.

We raised $494 with the groupā€™s Gitcoin grant Gitcoin DAO Anti-Fraud Workstream: Round 10 | Grants | Gitcoin

The funding will be allocated to those who actively participated in the Evaluation Panel in Round 10 and supported working group operations.

A huge THANK YOU! :pray: for your comments and interest, and to everyone who participated & contributed in Round 10!

6 Likes

Greetings humans! I wanted to create a workstream on Moving to the price of forgery, but just noticed this one. Think itā€™s better to publish here.

Moving to the price of forgery + turning gitcoin into identity provider.

Quality control of human verification methods

(or the way we assign trust bonuses to verification methods)

Systems like Proof of Humanity, Bright ID, Idena, POAP and ENS are already doing an amazing job in moving towards a no-bots future. Yet thereā€™s room for improvement. Which is better quality control and a fairer trust bonus for each method as a result.

@Adamscochran proposed to create a standard assessment procedure for new (and existing) methods. The procedure is incredibly detailed and exhaustive. It can describe literally any existing verification method quantitatively. Yet as well as the current procedure it isā€¦

ā€¦subjective. we cannot predict the ingenuity of malicious actors. Thereā€™s no way we can think of all the exploits for a particular verification method. And thereā€™s no simple way we can tell the method was exploited. An exploited verification method + exploited quality control algorithm is a Trojan horse for a bot army. So the first problem is the high risk to be a step behind malicious actors.

ā€¦and speculative. See how @Adamscochran scored Idena. It is so low that Idena would be kicked out of the approved methods. Iā€™d disagree with that. I love Idena and most probably would propose a different scoring for it or probably would demand to reconsider weights in scoring parameters. But we cannot really determine which weights or scores are right. Thus the second problem is the lack of objective measurement. Without it, we leave room for political games, biased opinions, and unproductive disputes (same as disputes on religion, politics, vaccines;)).

How price of forgery works

The solution to both of the problems described is the price of forgery concept.

The main idea is that the user score is valued in dollars and the user can grab these dollars at any time. After that, the userā€™s ID is deleted forever. The procedure is called ā€œexplosionā€ in Upala terms (we invented the concept and building the protocol implementing it). Iā€™ll use the term further.

With the price of forgery approach, we donā€™t have to rely on any assumptions about any verification method. The market will drive the scoring of each method to its price of forgery organically. We also get an assurance for future exploits. If a method got hacked we would see it immediately through an increased explosion rate. That would mean the method should be fixed, or its score should be decreased.

How Gitcoin may benefit

Price of forgery gives the ability to quality control human verification methods with market forces. It is unbiased, precise, and responsive to new exploits. For the verification methods It would mean:

  • Fairer trust bonuses
  • Automated, transparent, and ongoing quality control
  • Incentive for third parties to bring new verification methods or invent new ones

The benefits from shifting to the price of forgery go a bit further than ā€œjustā€ quality control. Gitcoin has tremendous data on interactions on the platform. It is then possible to aggregate scores from multiple sources for a single user (which trust bonuses do), and also append an additional trust bonus to award ā€œproperā€ internal interaction. That would benefit users and will help assess own algorithms (help understand better what the ā€œproperā€ behavior is).

And even further. With all that data and algorithms measured with the price of forgery Gitcoin may start providing its own digested scores to other DApps and earn on that (Upala protocol allows to do that). Due to the high user base, it may become a starting point for a decentralized identity system of the future.

Work done, next steps, and coordination

You can track progress on developing the price of forgery protocol here (looking for contributors by the way - weā€™ve got Ampere-hours!).

I had a talk with @owocki about the price of forgery back in September 2020. Then described integration in this github issue. I learned recently that Gitcoin is continuing on the idea - thread. So seems like the Upalaā€™s price of forgery concept is already approved. Though I havenā€™t had any conversation with the gitcoin team on this and I donā€™t see any coordination. As Gitcoin is decentralized now I thought it is a good idea to start the discussion here.

Related topics:

Discussion

  • How do you feel about turning from manual assessment of verification methods to the price of forgery?
  • What are your thoughts on adding identity provider functionality to Gitcoin?
  • To Gitcoin team: is the price of forgery really on the roadmap?

I suggest recruiting administrators to ensure that every month or quarter, the account requires human certification. Managers conduct manual reviews at specific times

1 Like

I didnā€™t get the notifications and missed the calls, Iā€™ll try to watch the videos asap. How can I make sure I get notified of the next one? I added my e-mail and telegram to the spreadsheet.

Next Monthly Community Call tomorrow Add to your calendar here: AddEvent

Full workstream calendar here: AddEvent.com - Sign in

3 Likes

This workstream still active? I filled out the google sheet with my info. Been working with ML algorithms the last 5 years with a background in maths and stats. Happy to get involved

1 Like

Hi @stephenwilliams1990,

Thanks for jumping in, yes this workstream is active but with selective skills and roles

I will review your application.

1 Like

Iā€™d love to learn about this so I can contribute, I know almost nothing about it! Good luckā€¦

1 Like

Hey @KriptoHayat, glad to see your enthusiasm. You can apply to the Gitcoin Dao at this address: Gitcoin DAO Onboarding

The new and growing Gitcoin Dao is slowly taking more and more responsibilities. It is possible in the future the Dao will take over the whole grants program and more.

1 Like