17

I'm search for a tool that could compare source codes for similarity.

We have a very trivial system right now that has huge amount of false positives and the real positives can easily get buried in them.

My requirements are:

  • reasonably small amount of false positives
  • good detection rate (yeah these are going against each other)
  • ideally with a more complex output than just a single value
  • usable for C (C99) and C++ (C++03 and optimally C++11)
  • still maintained
  • usable for comparing two source files against each other
  • usable in non-interactive mode

EDIT:

To avoid confusion, the following two code snippets are identical and should be detected as such:

for (int i = 0; i < 10; i++) { bla; }

int i; while (i < 10) { bla; i++; }

The same here:

int x = 10; y = x + 5;

int a = 10; y = a + 5;

Yavar
  • 11,883
  • 5
  • 32
  • 63
Šimon Tóth
  • 35,456
  • 20
  • 106
  • 151
  • 1
    Does diff satisfy some of your needs ? – High Performance Mark Jun 06 '12 at 10:28
  • 1
    @HighPerformanceMark LOL, no it doesn't. – Šimon Tóth Jun 06 '12 at 10:29
  • 2
    it looks as if you want a tool that compares the semantic similarity of 2 codes, rather than the syntactic. I suspect that this is an unsolved problem and that SO will struggle to find a satisfactory solution for you. I think that, as written, your original question will mislead a lot of readers just as it misled me. I also think you should rewrite it completely to state clearly what sorts of comparison you are interested in. – High Performance Mark Jun 06 '12 at 10:36
  • 4
    What about Moss? http://theory.stanford.edu/~aiken/moss/ – Mike Kwan Jun 06 '12 at 10:39

6 Answers6

18

I've used MOSS in the past: http://theory.stanford.edu/~aiken/moss/ to detect plagiarized code. Since it works on a semantic level, it will detect the situations you presented above. The tool is language-aware, so comments are not considered in the analysis, and it goes a long way in detecting code that has been modified through simple search-and-replace of variable and/or function names.

Note: I used the tool a few years ago when I taught computer science in grad school, and it worked wonderfully in detecting code that had been yanked from the internet. Here is a well-documented account of similar application: http://fie2012.org/sites/fie2012.org/history/fie99/papers/1110.pdf

If you google "measure software similarity", you should find a few more useful hits: http://web.archive.org/web/20150219121637/http://www.ics.heacademy.ac.uk/resources/assessment/plagiarism/detectiontools_sourcecode.html

Jan
  • 2,178
  • 3
  • 14
  • 26
Throwback1986
  • 5,887
  • 1
  • 30
  • 22
7

Your problem in Computer Science Terminology maybe stated as Source Code Plagiarism Detection. A good start would be to read this article on Dr Dobbs: Detecting Source-Code Plagiarism. It lists the Algorithms for detecting Plagiarism in the source code.

Note: What you have asked for is indeed a tough computing problem :)

Yavar
  • 11,883
  • 5
  • 32
  • 63
3

May be Copy-paste-detector from PMD?

Kirill
  • 46
  • 1
2

I start to use JPLAG (https://github.com/jplag/jplag) to check code similarity and compare students works in Java and text files. It works well to check same code structure and variable Substitution.

Pierre Jean
  • 366
  • 2
  • 9
1

You could try duplo. It will find common lines. It has some ability to ignore whitespace changes, but doesn't detect code with renamed variables, so it is more a cleanup-aid than a help when detecting plagiarism.

Benjamin Bannier
  • 55,163
  • 11
  • 60
  • 80
1

(response is late, but the question's relevance never goes away)

I was faced a similar problem and wrote a web based application.

https://jefferey-cave.gitlab.io/miss/

I was teaching in javascript and python, so those are the languages it handles. It does not handle C/C++ (currently). I'd be curious to see how the Javascript interpreter handles C.

available on gitlab

The problem I was faced with was it being illegal to submit student code across international boundaries (MOSS was forbidden) so needed something that would run locally. The implementation is pure client-side browser.

Interestingly, I think it may satisfy your "low false positive" criteria in a unique way. There is always some similarity between the assignments, but uses a unique technique to identify the point things "start to look suspicious".

I found it more useful in determining group dynamics in the classroom (who is working/studying with whom).

It has some fun live graphics, so it was useful to show to an Undergrad class after they submitted their first assignment. There was always a high degree of similarity in the first assignment, so no harm in demonstrating it live (with the submission names anonymized).

I always tell the story of the student I thought was (grossly and blatantly) cheating. Their work showed remarkable similarity to another student's very unique answer. Comparing the student's work to the rest of the class showed no significant similarity relative to the rest of the class. It was well below the threshold of "interesting". This led to a deeper investigation of the submission ... turns out there had been an tutorial, and the style showed through, but the work was unique.

Nothing happened, and those students never found out how close they came.

Jefferey Cave
  • 2,507
  • 1
  • 27
  • 46