The ability to detect plagiarism in undergraduate computer science classes is crucial, not only to promote fairness, but to discourage cheating among students. There are some frameworks that exist, but their effectiveness and ability to accurately group individuals who have plagiarized code may be limited. We have developed a graph-theoretical method for automated plagiarism detection in programming courses. The approach leverages code similarities detected by MOSS, a standard tool for performing pairwise code similarity comparisons. The method is primarily effective on programming assignments where the tasks are broken into smaller independent problems, as it can establish the relationships between the series of problems.