virtualphoton / mayat

Experimental AST-Based Source Code Similarity Detection Tool

Mayat

Mayat is a code similarity detection tool developed by Tian(Maxwell) Yang. It works by comparing the Abstract Syntax Trees of students' code solutions and generate a similarity score for each pair of students' code.

Build & Install

1. Clone the repo

git clone [email protected]:AnubisLMS/Mayat.git

2. Install dependencies

cd Mayat
pip install -r requirements_dev.txt

Install Mayat

pip install .

3. Install tree-sitter parsers

python -m mayat.install_langs

Usage

Let's say we need to check all students' uniq.c for homework1. The path for each uniq.c has the format homework1/<unique-id>/user/uniq.c. All we need to do is run:

python -m mayat.frontends.TS_C homework1/*/user/uniq.c

If we only want to check the main function, we can do:

python -m mayat.frontends.TS_C homework1/*/user/uniq.c -f main

Additionally, we can pass more optional arguments for C.py:

• --threshold: Specify the granularity for the matching algorithm. Default to 5. A smaller value will cause it to check trivial details, which increases the similarity score of two code even though they might not be similar. A larger value will cause it to overlook some common cheat tricks such as swapping two function definitions.

Supported Languages

• C:
  • mayat.TS_C
  • mayat.C(Legacy)
• Python:
  • mayat.TS_Python
  • mayat.Python(Legacy)
• Java:
  • mayat.TS_Java

Implement a New PL's frontend

We implement a new programming language's frontend by using classes and functions defined in mayat. They are:

• mayat.AST.AST: The base class for Abstract Syntax Tree. For a new PL you should inherit this and implement the AST.create(path) class method, which takes the path of a program as a parameter and returns the AST representation of that program. Currently it is preferred to use tree-sitter parsers to implement language frontends, whose corresponding file should be prefixed with TS_.
• mayat.args.arg_parser: A argparse.ArgumentParser object. We need to use this object to retrieve command arguments. We can add new arguments if needed.
• mayat.driver.driver: The driver function that takes the inherited AST class and the parsed arguments as parameters and run the plagiarism detection algorithm.

An example of this can be find in mayat/frontends/TS_C.py, which is a C frontend implemented using tree-sitter-c parser.

Testing

cd tests
python test.py -v

Limitations

This tool will never work for assembly code as the code has to be written in a high level programming language that can be converted into an AST. We can potentially figure out a way to automatically reverse engineer assembly code back to C and then convert it to AST. However, there's no guarantee that the reverse-engineered code can be a good representation for its assembly counterpart.