pyjectify 0.5

A Python library for memory manipulation, code injection and function hooking

PyJectify

A Python library for memory manipulation, code injection and function hooking.

Quick start

PyJectify is available on PyPI.

Alternatively, you can download releases from GitHub or clone the project.

Documentation is available at https://petitoto.github.io/pyjectify/

Features

Windows (x86 & x86_64)

Core

• Allocate / Free / Read / Write memory
• Create threads
• List loaded modules
• PE parser
• Use kernel32 or ntdll functions

Modules

• MemScan: scan memory using regex patterns
• Inject: load library, from disk (remote LoadLibrary) or from memory (fully map the DLL into the remote process)
• Hook: set up inline hooks in the target process
• PythonLib: embed python into a remote process (Python 3.6 - 3.11 supported)

Utils

• Syscall: parse syscall codes from ntdll.dll (from the loaded library or from the disk), and produce a ntdll-like object which can be used by the core to use direct syscalls
• ApiSetSchema: parse Windows ApiSet

Examples

Memory search & basic operations

import pyjectify

# Open notepad process (only the first found if multiple instances of notepad are running)
notepad = pyjectify.byName('Notepad.exe')[0]

# Use the pattern "secret( is)?: (.){10}", but encoded in utf-16-le because Notepad uses wchar_t
words = ['secret', ' is', ': ', '.']
pattern = b'%b(%b)?%b(%b){10}' % tuple(e.encode('utf-16-le') for e in words)

# Search for the secret in notepad's memory
addrs = notepad.memscan.scan(pattern)

# Process found addresses
for addr in addrs:
    secret = notepad.process.read(addr, 50).decode('utf-16-le')
    print('[+] Found secret:', secret)
    notepad.process.write(addr, ('*'*len(secret)).encode('utf-16-le')) # let's hide the secret!

# Reset memscan to perform a new search regardless of the previous scan
notepad.memscan.reset()

Python code injection

import pyjectify

# Open notepad process
notepad = pyjectify.byName('Notepad.exe')[0]

# Inject Python DLL
notepad.pythonlib.python_mod = notepad.inject.load_library("C:\\path\\to\\python-embed\\python311.dll")

# Run some Python code from notepad
notepad.pythonlib.initialize()
notepad.pythonlib.exec('import os; os.system("calc.exe")')

# Undo all initializations
notepad.pythonlib.finalize()

Setup an inline hook written in Python

import pyjectify

# Open notepad process & inject Python DLL
notepad = pyjectify.byName('Notepad.exe')[0]
notepad.pythonlib.python_mod = notepad.inject.load_library("C:\\path\\to\\python-embed\\python311.dll")
notepad.pythonlib.initialize()

# Let's hook GetClipboardData!
# Step 1: define our new function
pycode = """
import ctypes
def GetClipboardData(uFormat:ctypes.c_uint) -> ctypes.c_void_p:
  ctypes.windll.user32.MessageBoxW(0, "I hooked you :D", "MyNewGetClipboardData", 0)
  return o_GetClipboardData(uFormat)
"""
notepad.pythonlib.exec(pycode)

# Step 2: get original function address and setup a trampoline (of 15 bytes size)
user32 = notepad.process.get_module('user32.dll')
oaddr = user32.exports['GetClipboardData'] + user32.base_addr
trampoline_addr = notepad.hook.trampoline(oaddr, 15)

# Step 3: prepare Python function hooking, ie create o_GetClipboardData and get ou Python GetClipboardData address
hook_addr = notepad.pythonlib.prepare_hook('GetClipboardData', trampoline_addr)

# Step 4: inline hook
notepad.hook.inline(oaddr, hook_addr)

Advanced DLL injection

import pyjectify

# Open processes
proc1 = pyjectify.byName('proc1.exe')[0]
proc2 = pyjectify.byName('proc2.exe')[0]

# Extract a library from proc1's memory
module = proc1.process.get_module('module.dll')[0]

# Extract common syscalls from ntdll.dll and wrap them into a ntdll-like object
syscall = pyjectify.windows.Syscall()
syscall.get_common(from_disk=True)

# Use direct syscalls to operate on proc2 (memory read / write / protect, thread creation...)
proc2.process.ntdll = syscall

# Inject the module directly from memory into proc2, at a random location and without copying PE headers
proc2.inject.memory_loader(module, prefer_base_addr=False, copy_headers=False)