Assertions are a useful developer tool, but they cannot be relied upon to be applied in a production environment. Only use assert for internal invariants and debug-only checks that are not required for correct or safe execution.

Do not use assert for:

  • Security checks (including input validation)
  • Enforcing runtime conditions whose removal would affect correctness or security
  • Error handling

Reason:

Python’s optimization options (-O and -OO) remove assert statements from the bytecode and set __debug__ to False [Python 3.9 Documentation]. This means any logic implemented using assert may not execute in optimized runs.

As a result, using assert incorrectly (that is, for security or correctness-critical checks) can introduce vulnerabilities or other undesired behavior when optimization is enabled.

Instead, raise appropriate exceptions to ensure the checks are always enforced regardless of interpreter settings.

Non-Compliant Code Example

The non-compliant code example is checking for invalid arguments by using assertions. In this example, any positive integer between 1-709 inclusive is valid, and any other argument is invalid.

If the script is run normally, the assertions will catch the invalid arguments. If the script is run in optimized mode, assertions are removed from the bytecode and the function will not work as intended. To simplify the exploit code, the specific exception raised by the argument is caught.

noncompliant01.py:

""" Non-compliant Code Example """
import math


def my_exp(x):
    assert x in range(
        1, 710
    ), f"Argument {x} is not valid"  # range(1, 709) produces 1-708
    return math.exp(x)


#####################
# exploiting above code example
#####################

try:
    print(my_exp(1))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp(709))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp(710))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp(0))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp("b"))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

Output of noncompliant01.py:

$ python3.9 noncompliant01.py
2.718281828459045
8.218407461554972e+307
Argument 710 is not valid
Argument 0 is not valid
Argument b is not valid
$ python3.9 -O noncompliant01.py
2.718281828459045
8.218407461554972e+307
math range error
1.0
must be real number, not str

Compliant Solution

The my_exp() function raises a ValueError exception if an invalid argument is supplied. This works if the script is run in an optimized mode or not.

compliant01.py:

""" Compliant Code Example """
import math


def my_exp(x):
    if x not in range(1, 710):  # range(1, 709) produces 1-708
        raise ValueError(f"Argument {x} is not valid")
    return math.exp(x)


#####################
# exploiting above code example
#####################

try:
    print(my_exp(1))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp(709))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp(710))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp(0))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

try:
    print(my_exp("b"))
except (AssertionError, OverflowError, TypeError, ValueError) as e:
    print(e)

Output of compliant01.py:

$ python3.9 compliant01.py
2.718281828459045
8.218407461554972e+307
Argument 710 is not valid
Argument 0 is not valid
Argument b is not valid
$ python3.9 -O compliant01.py
2.718281828459045
8.218407461554972e+307
Argument 710 is not valid
Argument 0 is not valid
Argument b is not valid

Automated Detection

Tool Version Checker Description
Bandit 1.6.2 B101:assert_used Use of assert detected. The enclosed code will be removed when compiling to optimised byte code.
   
MITRE CWE Pillar
CWE-691: Insufficient Control Flow Management (4.13) (mitre.org)
MITRE CWE Base:
CWE-617, Reachable Assertion

Bibliography

   
[Python 3.9 Documentation] Python Software Foundation. (2024). Command line and environment - cmdoption -o [online].
Available from: https://docs.python.org/3.9/using/cmdline.html?highlight=pythonoptimize#cmdoption-o
[accessed 10 October 2024].