6

I have many public repositories working on projects related to machine learning stuff, and it has recently come to my notice that someone has published a machine learning research paper based on one of the projects that I have in my GitHub with the same results that would be obtained from running my code(without citing my GitHub, or taking prior permission).

I'm pretty upset. What can I do about it, from an academic route? And what could happen to the fraudster, if I do that?

6
  • 6
    First you need to assure yourself that it isn't just a case of parallel research.
    – Buffy
    May 29, 2020 at 20:45
  • 5
    Can you prove that they used your code? If so, then write to the editors/chairmen of the journal/conference with your proof and demand credit or retraction. May 29, 2020 at 20:49
  • 1
    The question could be improved by clarifying the nature of the projects and their use in the paper. For example, if you implemented some standard ML techniques, it wouldn't be surprising that someone else gets the same results as those produced by your code. May 30, 2020 at 6:40
  • 1
    What is the outcome you hope for? Assuming your code was used, asking for a citation might be easier than asking for the paper to be revised/withdrawn.
    – GoodDeeds
    May 30, 2020 at 8:15
  • 1
    @Enne You said that they produced results based on your methods. How do you know? If I invert a matrix I'll get the same results no matter whose code I use as long as it is matrix inversion. A couple of times, many years after I published something, somebody else (and more famous) did something very similar, but I don't doubt it was independent. A cool and actually quite natural idea, they didn't know about my work and thought they freshly invented the idea. This does happen. Of course, I could have argued with them about citing my old work, but frankly, I prefer to concentrate on my new one. May 30, 2020 at 12:28

1 Answer 1

5

The first question you need to ask is what license is your code under? If, for example, you published your code under the GPL, then someone taking your code and publishing results obtained with it is entirely within their rights. It may be bad academic style to not say where they got the code from, but there is nothing that would objectionable from a legal perspective about the fact. You might be upset that they didn't acknowledge you, but you likely have little recourse -- lots of people publish research done with widely used software packages like PyTorch without referencing where they had it from, and the academic community appears to be ok with that. (Whether that's how it should be is a separate question not to be litigated here.)

On the other hand, if you didn't attach a license to your code, then it remains yours whether or not the code is available on github. If you can prove that they used your code, then you could presumably take legal action against them, though that is a matter from which the journal in question will likely want to stay about as far away as possible. The difficult thing to show in this legal argument is that indeed the person in question did use your code, and not just their own implementation of a similar algorithm. If you have good reasons to believe that they did indeed use your code, then you will get a chance to test your belief during the discovery phase of any trial -- but you already get the idea: This is likely going to get expensive.

What would happen to that other person: If your pockets are deep enough to actually win a law suit, then that other person's employer is likely going to be interested in following up because their employee used a piece of software illegally. If your software was available under a license that allowed the other person to use it, then people can have differing opinions on whether or not they were required to cite you, and nothing will likely happen.

Short story short: If you really really don't want anyone to use your software, you have two options: (i) Use github, make your repositories public, mark everything as proprietary software, and set aside a few $10k for potential law suits; (ii) don't use github, or make your github repositories private.

5
  • I don't have the money for a legal battle. Would it not be possible to email them my code and reasoning and get the editors to contact the fraudster's institution, like @CaptainEmacs suggested?
    – Enne
    May 30, 2020 at 4:25
  • 1
    @Enne you have possible options, now decide what you will do.
    – Solar Mike
    May 30, 2020 at 6:38
  • Good answer, +1. One thing that might be relevant: if OP's code implements a unique algorithm which is mirrored in the paper, then the issue becomes one not of copyright, but of IP (which, if I make an educated guess here, is probably not protected by a patent by the OP). In this case, the OP still has the option of alerting the editors about plagiarism, which is not a legal, but an academic category of misdeed. I wonder if that may/should be added to the answer. May 30, 2020 at 12:31
  • @CaptainEmacs Correct -- it has nothing to do with copyright but with IP. What kind of IP you have is determined with whether or not a license was attached to the code and if so, which one. May 30, 2020 at 15:03
  • @Enne Yes, of course you can go email the author, the editors, anyone else you want, and complain. But what effect is that likely going to have? If I were the editor of the journal, given all I know about the situation, I'd just shrug and move on. If I were the department head of that person, given all I know about the situation, I'd ask the person next time we have lunch, I would decide that I don't know enough about who's right and who's wrong, shrug, and move on. The only thing it would achieve is that you might feel better. If that's what you're looking for, go for it! May 30, 2020 at 15:06

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .