przemub / tainter

Taint analysis for Python 3

Tainter

Taint analysis for Python 3

(Very early) work-in-progress. Here be dragons.

Taint analysis is a type of a static analysis of source code.

It is used to analyse the flow of information in the code, so we can make sure that there are no unintended leaks.

For example, let's say we store some data about the user, and we decided to use their insurance number as an ID:

class User:
    insurance_number: str
    date_of_birth: datetime

    @property
    def user_id(self):
        return self.date_of_birth + self.insurance_number

Sometime down the line, a well-meaning project manager uses that user ID in a public report, not knowing that it is generated from personally identifiable information. Time to lawyer up!

A taint analyser may use an annotation like this:

class User:
    @mark_tainted
    insurance_number: str
    @mark_tainted
    date_of_birth: datetime

    @property
    def user_id(self):
        return self.date_of_birth + self.insurance_number

And figure out that user_id is tainted as well, throwing a huge warning whenever it's used, until we rewrite it to be safe and mark it as such:

class User:
    @mark_tainted
    insurance_number: str
    @mark_tainted
    date_of_birth: datetime

    @mark_safe
    @property
    def user_id(self):
        return hash(self.date_of_birth + self.insurance_number + SALT)

Usage

Install

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Use

usage: main.py [-h] [-d] [-t TAINTED_ARG [TAINTED_ARG ...]] filename function

Taint analysis is a type of a static analysis of source code.

It is used to analyse the flow of information in the code, so
we can make sure that there are no unintended leaks.

For starters, check out the examples files to see how taint spreads.

Then run the tainter on any of the examples, for example:
python3 main.py --tainted-arg a examples/assignment.py chain
to see how the return value is tainted by input "a" in function chain
in file assingment.py.

positional arguments:
  filename              File to analyse.
  function              Function to analyse.

optional arguments:
  -h, --help            show this help message and exit
  -d, --dump            If selected, dump the AST for the function.
  -t TAINTED_ARG [TAINTED_ARG ...], --tainted-arg TAINTED_ARG [TAINTED_ARG ...]
                        Mark a function argument as tainted.

Test

examples is full of test cases. Run python -m unittest to run all of them.