Entropy Based Verification Of Academic Integrity

The usage of online submission provides an efficient means of facilitating course components; especially those with large enrolment. However, this convenience is not without caveat as student solutions are then as easily distributed as they submitted for credit. While services exist to combat academic dishonesty, such as online comparison databases, privacy concerns have been raised about their usage. We have developed and implemented an entropy based method for the verification of academic integrity. This tool was implemented primarily for a freshman undergraduate programming course with a yearly enrolment of more than 1000 students. Even with significant resources and effort dedicated to ensuring academic integrity, the result was difficult to quantify. Further compounding the challenge was the fact that computer program source code has rigidly defined constructs and syntax, thus a simple text comparison could indicate a high level of similarity that might implying a lower level of integrity. Conversely, two logically identical programs could be written with different variable names where a simple text comparison could indicate a higher level of integrity. Rather than performing a straight comparison, our entropy based method generates a symbolic library of the file and then analyses the library structure against all other file libraries as a measure of academic integrity; this method defeats the short comings of the aforementioned methods. This paper will present our entropy based method and its high level of success verifying the academic integrity of large sets of assignment submissions.