Secure, Offline Feedback to Convey Instructor Intent

One of the core skills that Computer Science undergraduates master is to create finite automata and regular expressions from natural-language descriptions of formal languages. It is quite common for such descriptions to be ambiguous (or appear to be so), particularly when instructors or students are not fluent in the natural language. Two questions arise: (1) Can an instructor unambiguously convey the intended description to students in a secure manner (i.e., without revealing the instructor´s own solution)? (2) Can such a tool operate offline (i.e., without requiring internet access)? Question (1) on its own can be addressed by permitting students to upload their attempts to a server, where they can be checked against the instructor´s solution for equivalence. Such feedback cannot be obtained by students lacking reliable internet connectivity. In this paper, we present a technique to address both these questions: assuming cryptographically secure hash functions exist, a student learns only whether her attempt is equivalent to the instructor´s solution (but nothing further), and after downloading the problem statement, all subsequent computation is performed on the student´s own computer. We have implemented this functionality as an open-source extension to JFLAP, a popular educational tool in this domain.