How MOSS Works
MOSS (Measure of Software Similarity) is a very effective plagiarism detection system that is commonly used by computer science professors across the country. Most students know that it’s nearly impossible to get away with plagiarism when MOSS is used, but not many know of how MOSS works or why it’s so effective. In this post, I will give an overview of copy-detection, document fingerprinting, and explain the winnowing algorithm used by MOSS. Then I will discuss the implementation and use of MOSS, and my experience evaluating the effectiveness of an OCaMOSS, an OCaml implementation of MOSS (my final project for CS 3110).
Note: This post is based on Winnowing: Local Algorithms for Document Fingerprinting and the writeup for OCaMOSS. All figures are borrowed from those 2 papers.
MOSS is a type of copy-detection algorithm. Given a set of documents, a copy-detection algorithm identifies pairs of documents which are likely to have copied from each other. There are three properties that are required in effective copy-detection algorithms:
