An Experience of Detecting Plagiarized Source Codes in Competitive Programming Contests

摘要

We proposed a new similarity measure for detecting code plagiarism based on the local alignment. The new measure is based on the gapped local alignment proposed by Smith and Waterman [2]. Our local alignment algorithm is an adaptive variant [1] of Smith-Waterman's. The novel feature of our similarity measure is in the normalization method. The similarity score of the aligned region is generally divided by the maximum possible score to get a normalized score. However, we adopted an exponential normalization function which is similar to Weibull distribution.