Machine Learning for identifying SCAM NFT Games

[Maus0728]

Hi everyone!

I just wanna ask if there's interest in an application that uses machine learning to identify potentially fraudulent NFT games. Machine Learning is basically an approach in which we feed raw data to an application/machine and it learns something on its own[1]. And I understand that there is a lot of data to examine, such as their website, github repository, whitepaper and fake team to be utilized as Artificial Intelligence training data, and so on...but is there even a demand for it? Is this a worthwhile research project?

On the other hand, I discovered a previous study[2] that utilizes the similar process but instead of NFT, they identify scam ICOs, which provides me an idea for this study's topic. With the NFT games at peak, will this topic be helpful in many ways?

[1] https://www.expert.ai/blog/machine-learning-definition/#:~:text=Machine%20learning%20is%20an%20application,it%20to%20learn%20for%20themselves.
[2] https://arxiv.org/abs/1803.03670

*Self moderated for unconstructive posts*

[1714871107]

1714871107

[1714871107]

1714871107

[jc12345]

Hi everyone!

I just wanna ask if there's interest in an application that uses machine learning to identify potentially fraudulent NFT games. Machine Learning is basically an approach in which we feed raw data to an application/machine and it learns something on its own[1]. And I understand that there is a lot of data to examine, such as their website, github repository, whitepaper and fake team to be utilized as Artificial Intelligence training data, and so on...but is there even a demand for it? Is this a worthwhile research project?

On the other hand, I discovered a previous study[2] that utilizes the similar process but instead of NFT, they identify scam ICOs, which provides me an idea for this study's topic. With the NFT games at peak, will this topic be helpful in many ways?

[1] https://www.expert.ai/blog/machine-learning-definition/#:~:text=Machine%20learning%20is%20an%20application,it%20to%20learn%20for%20themselves.
[2] https://arxiv.org/abs/1803.03670

*Self moderated for unconstructive posts*

Being able to tell if something is a scam with more certainty is really something that is needed. Although there are current indicators to guide us it is never possible to tell with 100% certainty and there is always a risk of something being a scam, but sometimes it only materializes after the scam occurred. An indicator you mentioned is having a whitepaper, but not having a whitepaper does not mean it is a scam, it just adds to the risk score.

Some important items to consider:

• It may be good to produce a risk score instead of claiming something is a scam until there are clear evidence. Factors that could impact a risk score are items you mentioned, as well as when contracts are involved the audit reports. A factor that will be very difficult to program is the gut feel of the idea. The idea may sound good but on closer inspection it seems at some point there will be division by zero proverbially speaking or lead turned into gold, that should count a lot. A particular issue for me is the project claims that there will be payouts in a stable coin but they don't say where the stable coin funds will come from.
• When is something considered to be a scam and when is it not - for example, dev does a rug pull or exist with the funds and the project died, funds were obtained but the dev team ran out of funds and the project died,  there was no community interest and a project died, some government shut something down and the project died, a dev team had a great idea but poor strategy and execution and the project died etc.
• What data will be used for he machine to learn eg will you use attributes from past confirmed scams?
• At which point do you confirm a project is a scam. There are projects that have been going on for years but is not progressing a lot with some people claiming scam and others clinging to hope

[crwth]

I think there's always a market for everything but it will depend on what's your purpose in creating it. What I mean by this is that is it for profit or is it for the betterment of the whole economic system to help identify scams and automate them?

I'm not well-vested with machine learning type of things but it sounds like a good idea if the cost of it is minimal to run for sure. I think it weighs more to help others identify scams and not just be worried about the cost of operating them.

Anyway, I tried looking at the article for the machine learning for IcoRating but can't seem to download a copy of their paper. I wanted to learn the introduction and conclusion that they have with it. How they handle false positives and true negatives and what are the percentage of it etc.

Can this be a good reference for this possible project?
https://github.com/awslabs/fraud-detection-using-machine-learning

[Maus0728]

An indicator you mentioned is having a whitepaper, but not having a whitepaper does not mean it is a scam, it just adds to the risk score.

Some important items to consider:
[...snip...]

Hi @jc12345 . Thank you for your constructive inputs. I'll make sure to take note all of the important aspect you've mentioned and I'll discuss this with my groupmates.

And, to be honest, based on your recommendations, identifying scam projects using machine learning appears to be quite difficult due to the large number of factors that must be taken into account, and I agree that having some type of barometer or scale to measure the risk factor is more realistic than simply declaring whether a project is a scam or not. And, just so you know, this is just a research topic for my undergrad degree, and it's one of the few options we're considering.

Thank you very much.

I think there's always a market for everything but it will depend on what's your purpose in creating it. What I mean by this is that is it for profit or is it for the betterment of the whole economic system to help identify scams and automate them?

This is for my undergraduate research paper, but upon seeing all the recommendations given by @jc12345, it seems that it is impossible to create such system due to an extremely vast sets of data especially now that we are still on the introduction of Machine Learning without any prior experience of at least the basics. 

Anyway, I tried looking at the article for the machine learning for IcoRating but can't seem to download a copy of their paper. I wanted to learn the introduction and conclusion that they have with it. How they handle false positives and true negatives and what are the percentage of it etc.

I forgot to include the download link of the said research paper but here it is:
https://www.researchgate.net/publication/323722883_IcoRating_A_Deep-Learning_System_for_Scam_ICO_Identification

Can this be a good reference for this possible project?
https://github.com/awslabs/fraud-detection-using-machine-learning

Thanks for additional references. Very much appreciated.

[vv181]

is there even a demand for it?

One thing I could think of is the demand to create a web NFT-rating based on machine learning may be plausible, not specifically about the games but in a broader sense, some people may want to know whether some specific NFT is a copycat/legitimate or not. You can also build a system that detects whether some NFTs have been washed-trade or not. Though, neither I know much about machine learning, so I don't whether those things I mentioned are achievable or maybe overkill if you are using machine learning for it.

~Is this a worthwhile research project?

On the other hand, I discovered a previous study[2] that utilizes the similar process but instead of NFT, they identify scam ICOs, which provides me an idea for this study's topic. With the NFT games at peak, will this topic be helpful in many ways?

Looking at the referenced study, it has been cited over 31 times, so if your group are passionate enough, it may help other researchers specifically about NFT instead of ICO.

[Zackgeno96]

I would second on that statement that this topic has a vast scope, too broad which will make it extremely challenging, thus I suggest that if you would really want to pursue such then you should just limit the focus of your study, OP. Although having this topic will be like no other, it is important that the streamline of the scope of the study will be well communicated to your future readers and panels. After all, it wouldn’t be saving much of its purpose if other researchers and readers won’t be able to understand it.

[tbct_mt2]

Machine learning can help you to scan source codes and check that whether a source code of one project is similar or clone from other projects. After that, if you see very little activities of coding, developments in their Github, it would be signal of a low quality project. In turn, many low quality projects are from scammers and have unhappy endings.

Scam or not, it is not big matter if in the end, they are all failed projects. As investors, you will lose your capital if you invest in a project which ends as failed or scam.

[bitcoinnews_id]

Some important items to consider:

• It may be good to produce a risk score instead of claiming something is a scam until there are clear evidence. Factors that could impact a risk score are items you mentioned, as well as when contracts are involved the audit reports. A factor that will be very difficult to program is the gut feel of the idea. The idea may sound good but on closer inspection it seems at some point there will be division by zero proverbially speaking or lead turned into gold, that should count a lot. A particular issue for me is the project claims that there will be payouts in a stable coin but they don't say where the stable coin funds will come from.
• When is something considered to be a scam and when is it not - for example, dev does a rug pull or exist with the funds and the project died, funds were obtained but the dev team ran out of funds and the project died,  there was no community interest and a project died, some government shut something down and the project died, a dev team had a great idea but poor strategy and execution and the project died etc.
• What data will be used for he machine to learn eg will you use attributes from past confirmed scams?
• At which point do you confirm a project is a scam. There are projects that have been going on for years but is not progressing a lot with some people claiming scam and others clinging to hope

I totally agree with @jc12345 has said that I quoted. Let me add a little thing, that the project is more perfect in encapsulating the end goal of the creator. Indicators that will be studied must really have shown previous evidence. So that your claim to the results of your research is not ambiguous. You can test your research results with many methods such as confusion matrix to determine how accurate and precise it is. I know this will be very difficult to run, you have to find and collect data to create your own dataset because it is not available on the internet or libraries.

If this research can be done according to your hypothesis, then find a way and finish it. I hope you find new insights and contributions to technology and education.

[Maus0728]

Ey!! Thanks for all the replies!

Unfortunately, our professor rejected this research idea, possibly because he isn't interested in this NFT thingy. But, given the complexity of the research and our lack of knowledge with various machine learning algorithms, we're fine with his decision.

Nonetheless, the previously given pointers are something worth experimenting with as soon as we have the chance outside academics.

[asyakashi]

On the other hand, I discovered a previous study[2] that utilizes the similar process but instead of NFT, they identify scam ICOs, which provides me an idea for this study's topic. With the NFT games at peak, will this topic be helpful in many ways?

I am sure this topic will help a lot of people especially in the future blockchain programmer field.
We really need a tool to identify NFT scams, if necessary anything related to new cryptocurrencies.
But this is machine learning, of course not everyone can understand it, usually they use reliable programmers to take advantage of this technology.
With this topic of yours, I'm sure it's been enough to help a lot of people to believe by not worrying about NFT scams, we also need to keep updating about the tools you share so that programmers can use this as well as possible.

[el kaka22]

Machine learning is the tech revolution of our time. We are talking about a software that reads some stuff and learns it and gets better at doing it. First time, I ever heard of it was chess, because in chess a machine learning software could learn how things move and get better at it and basically they reached to a level where even the best in the world could not beat it.

So, is there a potential for such a software to exist, then yes there is. However, it is not about software being used or not, obviously everyone would want to avoid scam NFT projects and would pay good money to do so, however you have to do marketing to both get yourself heard, AND also convince people it actually works.