Plagiarizing and Paraphrasing Code from an Online Class for Content Marketing | Hacker News Reader

Plagiarizing and Paraphrasing Code from an Online Class for Content Marketing

minimaxir.com

34 pointsminimaxir10 comments

> But the willful and unnecessary paraphrasing of code, replacing all the CamelCase variables with underscored variables, and code which is worse than the original, is hard to overlook. “I recommend you for example this edX course” does not give sufficient credit to the teachers and staff of CS100.1x.

Anyone here have access to one of those code-plagiarizing detection systems used at CS schools? How would it fare on this kind of obfuscation?

I know that a plagiarized program that tries to just futz variable names is probably computationally easy to detect. And even extremely computationally exhaustive comparisons would be feasible, because the average college only has to deal with a few hundred to a thousand students a semester.

But would the algorithm even have to be more sophisticated than doing just a plain diff and applying some heuristic, even just the ratio of identical lines to total lines? My perception of students who would cheat in a 101CS class is that they are so incompetent that they don't even know how to change a variable name in a program without breaking the code...nevermind rearranging the order of a loop operation. Any student who could confidently obfuscate their plagiarism of 101-level program assignment is probably smart enough to just do the program themselves.

In the case of CS100.1X, you had to submit the program to an online grader. However, the assignment contains test cases; if all your test cases pass, then you'd most likely pass the online grader.

However, if you changed the variable names, like in the article, the test cases would fail...unless you changed them in the test cases itself. In which case, the test cases would definitely fail in the online grader.

I took the Andrew Ng Coursera class which had an online grader that had test cases as well, but mostly it was just defining functions with specified input and outputs, so you're free to use whatever style guide you want in between. (CS100.1X does have a few of those, and they are easily the most difficult part of the assignment since it requires Python shenanigans)

As the name suggests, MOSS (Measure of Software Similarity) and other tools employed for plagiarism detection work by detecting logical similarity. Obfuscated code is considered identical to the original.

See: http://theory.stanford.edu/~aiken/publications/papers/sigmod...

At the Royal Institute of Technology (KTH), Stockholm a tool was developed called KATTIS (https://kth.kattis.com/) that not only check submitted code for plagiarism but also verifies it with test input/output as well as measuring execution time/memory usage.

There is a paper with more details "Five Years with Kattis - Using an Automated Assessment System in Teaching", https://www.csc.kth.se/~gkreitz/kattis-fie11/

The system is still developed and there is now an open version of it with a huge problem archive (competitive programming style) used by several universities: https://open.kattis.com/

Trivia: KATTIS was originally very much developed by KTH's successful competitive programming team "Three-headed monkey", consisting of some PhD students (https://www.csc.kth.se/~gkreitz/) that did their thesis on stuff like streaming and encryption which in turn resulted in the company... Spotify.

On the other hand, false positives are a problem with more advanced techniques that ignore things like variable names, especially if assignments are narrow in scope (as course assignments mostly are) and have exacting requirements.

For example, there aren't very many sane ways to write a Hello World program.

> And more underscore variables?

Variable names and function names are supposed to use underscores and not Camel Case in Python. I wish people who are writing tutorials teaching people Python would at least glance at PEP-8.

I can't seem to get the page to load even though

    --- minimaxir.com ping statistics ---
    7 packets transmitted, 7 received, 0% packet loss, time 6223ms
    rtt min/avg/max/mdev = 443.694/691.180/969.051/196.041 ms

It should work fine after a reload or two.

I've been attempting to debug connection issues for a long time to no success. Somehow, curl works, but not dig, which baffles me.

I think it's because of the interaction between CloudFlare and GitHub Pages. If anyone has a solution, let me know.

I did the first of the Spark courses too. I'd second the recommendation, though I did it without knowing Python and I got through OK.

Is it possible that the author wrote this without mentioning the course to avoid giving away the answers to the course?

Alternate theory, there was never an obfuscation attempt. They simply ran it through a code formatter. Still, it's unattributed, so I am not sure it matters.

Reply on news.ycombinator.com