Ask HN: University checks for cheaters when delivering code assigments. How? | Hacker News Reader

Ask HN: University checks for cheaters when delivering code assigments. How?

4 pointsredxblood8 comments

My university checks for possible cheaters when delivering programming assignments. by inputing them to a program that gives back the probablitlity those two programs have code copyed of each other. How does the program work in essence?

View on HN

It may use something like Kolmogorov complexity https://en.wikipedia.org/wiki/Kolmogorov_complexity#Compress...

...where symbol matching is used along with variable names after removing all extra white space.

Although, I'm guessing your university uses PMD : http://pmd.sourceforge.net It's a fairly popular tool for detecting code issues (including copied segments).

Take a look at this:

http://theory.stanford.edu/~aiken/moss/

Specifically:

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

I'm no expert in this but I suppose just like how they check for plagiarism in essays.

They usually compare submissions with other submissions, past submissions and also against similar references on the internet in case a match is found on some website.

I guess... but there has to be something else. You can plagarize an essay, and finding that is kinda easy because you copy-paste word by word and so, it's obvious. But two programmers can have identical pieces of code without that meaning they copied each other, don't you think?

Not really; discounting extremely simple programs like print "Hello world", there are many possible stylistic differences and people will write things differently. One person will write for, another while; one i=0;j=0, another j=0;i=0; another will name variables differently; another will take two lines and make them into a procedure; one will write "if x then return y else return z", another "if x then return y; return z;", another "if not x then return z else return y" another "return x ? y : z" etc. If your assignment is at least 50 lines of code, compare it with someone else. You will see tons of differences.

Are you asking out of intellectual curiousity or a desire to beat the system?

Curiosity. If i was capable of beating the system, then i might as well do the assignment!

It's probably a combination of runtime analysis and code construction analysis.

Submissions with similar runtimes will use the same algorithm. If most students use the proper way, but a group of five students all, by "chance," pick an obscure dumb way (or a really obscure fast way), then they may be copying. Manual investigation would be triggered.

Look into how JavaScript minimizers/obfuscators work. They could be parsing your code down to a form where your variable names and comments don't matter at all. They're just analyzing and looking for coincident structure of programs.

Now, your administration is smart enough to not flag code where there's one way to do it (write "find a member of a list"), but if 10 people copy the same wrong way to do it, it would be a flag to investigate further.

Reply on news.ycombinator.com