def register_io_in(self):
in_ioctl = x64.MultipleInstr()
INPUT_BUFFER_SIZE = x64.mem('[RCX]')
INPUT_BUFFER_PORT = x64.mem('[RCX + 8]')
in_ioctl += x64.Cmp(self.IO_STACK_INPUT_BUFFER_LEN, 0x10) # size indicator / port
in_ioctl += x64.Jnz(":FAIL")
in_ioctl += x64.Mov('RCX', self.IO_STACK_INPUT_BUFFER)
in_ioctl += x64.Mov('RDX', INPUT_BUFFER_PORT)
in_ioctl += x64.Mov('RCX', INPUT_BUFFER_SIZE)
in_ioctl += x64.Cmp('RCX', 0x1)
in_ioctl += x64.Jnz(":OUT_2_OR_4")
in_ioctl += x64.In('AL', 'DX')
in_ioctl += x64.Jmp(':SUCCESS')
in_ioctl += x64.Label(":OUT_2_OR_4")
in_ioctl += x64.Cmp('RCX', 0x2)
in_ioctl += x64.Jnz(":OUT_4")
in_ioctl += x64.In('AX', 'DX')
in_ioctl += x64.Jmp(':SUCCESS')
in_ioctl += x64.Label(":OUT_4")
in_ioctl += x64.In('EAX', 'DX')
in_ioctl += x64.Label(":SUCCESS")
in_ioctl += x64.Mov('RDX', self.IRP_OUTPUT_BUFFER)
in_ioctl += x64.Mov(x64.mem('[RDX]'), 'RAX')
in_ioctl += x64.Xor('RAX', 'RAX')
in_ioctl += x64.Ret()
in_ioctl += x64.Label(":FAIL")
in_ioctl += x64.Mov('RAX', 0x0C000000D)
in_ioctl += x64.Ret()
self.upgrade_driver_add_new_ioctl_handler(DU_IN_IOCTL, in_ioctl.get_code())
def register_io_out(self):
out_ioctl = x64.MultipleInstr()
INPUT_BUFFER_SIZE = x64.mem('[RCX]')
INPUT_BUFFER_PORT = x64.mem('[RCX + 8]')
INPUT_BUFFER_VALUE = x64.mem('[RCX + 0x10]')
out_ioctl += x64.Cmp(self.IO_STACK_INPUT_BUFFER_LEN, 0x18) # size indicator / port / value
out_ioctl += x64.Jnz(":FAIL")
out_ioctl += x64.Mov('RCX', self.IO_STACK_INPUT_BUFFER)
out_ioctl += x64.Mov('RDX', INPUT_BUFFER_PORT)
out_ioctl += x64.Mov('RAX', INPUT_BUFFER_VALUE)
out_ioctl += x64.Mov('RCX', INPUT_BUFFER_SIZE)
out_ioctl += x64.Cmp('RCX', 0x1)
out_ioctl += x64.Jnz(":OUT_2_OR_4")
out_ioctl += x64.Out('DX', 'AL')
out_ioctl += x64.Jmp(':SUCCESS')
out_ioctl += x64.Label(":OUT_2_OR_4")
out_ioctl += x64.Cmp('RCX', 0x2)
out_ioctl += x64.Jnz(":OUT_4")
out_ioctl += x64.Out('DX', 'AX')
out_ioctl += x64.Jmp(':SUCCESS')
out_ioctl += x64.Label(":OUT_4")
out_ioctl += x64.Out('DX', 'EAX')
out_ioctl += x64.Label(":SUCCESS")
out_ioctl += x64.Xor('RAX', 'RAX')
out_ioctl += x64.Ret()
out_ioctl += x64.Label(":FAIL")
out_ioctl += x64.Mov('RAX', 0x0C000000D)
out_ioctl += x64.Ret()
self.upgrade_driver_add_new_ioctl_handler(DU_OUT_IOCTL, out_ioctl.get_code())
def sc_64_LoadLibrary(dll_path):
dll = bytes("KERNEL32.DLL\x00".encode("utf-16-le"))
api = b"LoadLibraryA\x00"
if PY3 and isinstance(dll_path, str):
dll_path = bytes(dll_path.encode())
LoadLibrary64_sc = x64.MultipleInstr()
LoadLibrary64_sc += shellcraft.amd64.pushstr(dll)
LoadLibrary64_sc += x64.Mov("RCX", "RSP")
LoadLibrary64_sc += shellcraft.amd64.pushstr(api)
LoadLibrary64_sc += x64.Mov("RDX", "RSP")
LoadLibrary64_sc += x64.Call(":FUNC_GETPROCADDRESS64")
LoadLibrary64_sc += x64.Mov("R10", "RAX")
LoadLibrary64_sc += shellcraft.amd64.pushstr(dll_path)
LoadLibrary64_sc += x64.Mov("RCX", "RSP")
LoadLibrary64_sc += x64.Sub("RSP", 0x30)
LoadLibrary64_sc += x64.And("RSP", -32)
LoadLibrary64_sc += x64.Call("R10")
LoadLibrary64_sc += x64.Label(":HERE")
LoadLibrary64_sc += x64.Jmp(":HERE")
LoadLibrary64_sc += windows.native_exec.nativeutils.GetProcAddress64
return LoadLibrary64_sc.get_code()
def register_alloc_memory(self):
ExAllocatePoolWithTag = self.kdbg.get_symbol_offset("nt!ExAllocatePoolWithTag")
if ExAllocatePoolWithTag is None:
raise ValueError("Could not resolve <ExAllocatePoolWithTag>")
INPUT_BUFFER_ALLOC_TYPE = x64.mem('[RCX]')
INPUT_BUFFER_ALLOC_SIZE = x64.mem('[RCX + 0x8]')
INPUT_BUFFER_ALLOC_TAG = x64.mem('[RCX + 0x10]')
Alloc_IOCTL = x64.MultipleInstr()
Alloc_IOCTL += x64.Cmp(self.IO_STACK_INPUT_BUFFER_LEN, 0x18)
Alloc_IOCTL += x64.Jnz(':FAIL')
Alloc_IOCTL += x64.Mov('RCX', self.IO_STACK_INPUT_BUFFER)
Alloc_IOCTL += x64.Mov('R8', INPUT_BUFFER_ALLOC_TAG)
Alloc_IOCTL += x64.Mov('RDX', INPUT_BUFFER_ALLOC_SIZE)
Alloc_IOCTL += x64.Mov('RCX', INPUT_BUFFER_ALLOC_TYPE)
Alloc_IOCTL += x64.Mov('RAX', ExAllocatePoolWithTag)
Alloc_IOCTL += x64.Call('RAX')
Alloc_IOCTL += x64.Mov('RBX', self.IRP_OUTPUT_BUFFER)
Alloc_IOCTL += x64.Mov(x64.mem('[RBX]'), 'RAX')
Alloc_IOCTL += x64.Xor('RAX', 'RAX')
Alloc_IOCTL += x64.Ret()
Alloc_IOCTL += x64.Label(":FAIL")
Alloc_IOCTL += x64.Mov('RAX', 0x0C000000D)
Alloc_IOCTL += x64.Ret()
self.upgrade_driver_add_new_ioctl_handler(DU_MEMALLOC_IOCTL, Alloc_IOCTL.get_code())
def sc_64_WinExec(exe):
dll = bytes("KERNEL32.DLL\x00".encode("utf-16-le"))
api = b"WinExec\x00"
if PY3 and isinstance(exe, str):
exe = bytes(exe.encode())
WinExec64_sc = x64.MultipleInstr()
WinExec64_sc += shellcraft.amd64.pushstr(dll)
WinExec64_sc += x64.Mov("RCX", "RSP")
WinExec64_sc += shellcraft.amd64.pushstr(api)
WinExec64_sc += x64.Mov("RDX", "RSP")
WinExec64_sc += x64.Call(":FUNC_GETPROCADDRESS64")
WinExec64_sc += x64.Mov("R10", "RAX")
WinExec64_sc += shellcraft.amd64.pushstr(exe)
WinExec64_sc += x64.Mov("RCX", "RSP")
WinExec64_sc += x64.Sub("RSP", 0x30)
WinExec64_sc += x64.And("RSP", -32)
WinExec64_sc += x64.Call("R10")
WinExec64_sc += x64.Label(":HERE")
WinExec64_sc += x64.Jmp(":HERE")
WinExec64_sc += windows.native_exec.nativeutils.GetProcAddress64 # Dirty infinite loop
#WinExec64_sc +=# x64.Ret(),
return WinExec64_sc.get_code()
def generate_64bits_execution_stub_from_syswow(x64shellcode):
"""shellcode must NOT end by a ret"""
current_process = windows.current_process
if not current_process.is_wow_64:
raise ValueError(
"Calling generate_64bits_execution_stub_from_syswow from non-syswow process"
)
transition64 = x64.MultipleInstr()
transition64 += x64.Call(":TOEXEC")
transition64 += x64.Mov("RDX", "RAX")
transition64 += x64.Shr("RDX", 32)
transition64 += x64.Retf32() # 32 bits return addr
transition64 += x64.Label(":TOEXEC")
x64shellcodeaddr = windows.current_process.allocator.write_code(
transition64.get_code() + x64shellcode)
transition = x86.MultipleInstr()
transition += x86.Call(CS_64bits, x64shellcodeaddr)
transition += x86.Ret()
stubaddr = windows.current_process.allocator.write_code(
transition.get_code())
exec_stub = ctypes.CFUNCTYPE(ULONG64)(stubaddr)
return exec_stub
def generate_64bits_execution_stub_from_syswow(x64shellcode):
"""shellcode must NOT end by a ret"""
current_process = windows.current_process
if not current_process.is_wow_64:
raise ValueError(
"Calling generate_64bits_execution_stub_from_syswow from non-syswow process"
)
transition64 = x64.MultipleInstr()
transition64 += x64.Call(":TOEXEC")
transition64 += x64.Mov("RDX", "RAX")
transition64 += x64.Shr("RDX", 32)
transition64 += x64.Retf32() # 32 bits return addr
transition64 += x64.Label(":TOEXEC")
x64shellcodeaddr = thread_state.allocator.write_code(
transition64.get_code() + x64shellcode)
transition = x86.MultipleInstr()
transition += x86.Call(CS_64bits, x64shellcodeaddr)
# Reset the SS segment selector.
# We need to do that due to a bug in AMD CPUs with RETF & SS
# https://github.com/hakril/PythonForWindows/issues/10
# http://blog.rewolf.pl/blog/?p=1484
transition += x86.Mov("ECX", "SS")
transition += x86.Mov("SS", "ECX")
transition += x86.Ret()
stubaddr = thread_state.allocator.write_code(transition.get_code())
exec_stub = ctypes.CFUNCTYPE(ULONG64)(stubaddr)
return exec_stub
def test_x64_multiple_instr_add_instr_and_str():
res = x64.MultipleInstr()
res += x64.Nop()
res += "ret; ret; label :offset_3; ret"
res += x64.Nop()
res += x64.Label(":offset_5")
assert res.get_code() == b"\x90\xc3\xc3\xc3\x90"
assert res.labels == {":offset_3": 3, ":offset_5": 5}
def test_get_context_address_64(self, proc64):
code = x64.MultipleInstr()
code += x64.Mov("RAX", 0x4242424243434343)
code += x64.Label(":LOOP")
code += x64.Jmp(":LOOP")
t = proc64.execute(code.get_code())
time.sleep(0.5)
cont = t.context
assert cont.Rax == 0x4242424243434343
def register_kernel_call(self):
# expect in buffer: the address to call and all dword to push on the stack
CCall_IOCTL = x64.MultipleInstr()
CCall_IOCTL += x64.Mov('RAX', self.IO_STACK_INPUT_BUFFER_LEN)
CCall_IOCTL += x64.Cmp('RAX', 0)
CCall_IOCTL += x64.Jz(":FAIL") # Need at least the function to call
CCall_IOCTL += x64.Mov('R15', 4 * 8) # Size to pop on the stack at the end (4 * push RDI)
CCall_IOCTL += x64.Mov('R10', self.IO_STACK_INPUT_BUFFER)
CCall_IOCTL += x64.Label(':PUSH_NEXT_ARG')
CCall_IOCTL += x64.Cmp('RAX', (8 * 5))
CCall_IOCTL += x64.Jbe(":SETUP_REG_ARGS")
CCall_IOCTL += x64.Sub('RAX', 8)
INPUT_BUFFER_NEXT_ARG = x64.create_displacement(base='R10', index='RAX')
CCall_IOCTL += x64.Mov('RBX', INPUT_BUFFER_NEXT_ARG)
CCall_IOCTL += x64.Push('RBX')
CCall_IOCTL += x64.Add('R15', 8) # Add at Size to pop on the stack at the end
CCall_IOCTL += x64.Jmp(':PUSH_NEXT_ARG')
CCall_IOCTL += x64.Label(":SETUP_REG_ARGS")
# Could be done in a loop
# But do I really want to generate x86 in a loop..
CCall_IOCTL += x64.Cmp('RAX', (8 * 5))
CCall_IOCTL += x64.Jz(":SETUP_4_ARGS")
CCall_IOCTL += x64.Cmp('RAX', (8 * 4))
CCall_IOCTL += x64.Jz(":SETUP_3_ARGS")
CCall_IOCTL += x64.Cmp('RAX', (8 * 3))
CCall_IOCTL += x64.Jz(":SETUP_2_ARGS")
CCall_IOCTL += x64.Cmp('RAX', (8 * 2))
CCall_IOCTL += x64.Jz(":SETUP_1_ARGS")
CCall_IOCTL += x64.Jmp(":SETUP_0_ARGS")
CCall_IOCTL += x64.Label(":SETUP_4_ARGS")
CCall_IOCTL += x64.Mov('R9', x64.mem('[R10 + 0x20]'))
CCall_IOCTL += x64.Label(":SETUP_3_ARGS")
CCall_IOCTL += x64.Mov('R8', x64.mem('[R10 + 0x18]'))
CCall_IOCTL += x64.Label(":SETUP_2_ARGS")
CCall_IOCTL += x64.Mov('RDX', x64.mem('[R10 + 0x10]'))
CCall_IOCTL += x64.Label(":SETUP_1_ARGS")
CCall_IOCTL += x64.Mov('RCX', x64.mem('[R10 + 8]'))
CCall_IOCTL += x64.Label(":SETUP_0_ARGS")
CCall_IOCTL += x64.Mov('RAX', x64.mem('[R10]'))
# Fix Reserve space (calling convention)
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Call('RAX')
CCall_IOCTL += x64.Mov('RDX', self.IRP_OUTPUT_BUFFER)
CCall_IOCTL += x64.Mov(x64.mem('[RDX]'), 'RAX')
CCall_IOCTL += x64.Xor('RAX', 'RAX')
CCall_IOCTL += x64.Add('RSP', 'R15')
CCall_IOCTL += x64.Ret()
CCall_IOCTL += x64.Label(":FAIL")
CCall_IOCTL += x64.Mov('RAX', 0x0C000000D)
CCall_IOCTL += x64.Ret()
self.upgrade_driver_add_new_ioctl_handler(DU_KCALL_IOCTL, CCall_IOCTL.get_code())
def test_get_context_address_64(self):
with Calc64() as calc:
code = x64.MultipleInstr()
code += x64.Mov("RAX", 0x4242424243434343)
code += x64.Label(":LOOP")
code += x64.Jmp(":LOOP")
t = calc.execute(code.get_code())
time.sleep(0.5)
cont = t.context
self.assertEqual(cont.Rax, 0x4242424243434343)
def _upgrade_driver_inject_base_upgrade(self):
kldbgdrv = self.kldbgdrv
upgrade = x64.MultipleInstr()
# R14 : IOCODE
# RSI -> IO_STACK_LOCATION
# RDI -> IRP
upgrade = x64.MultipleInstr()
upgrade += x64.Cmp('R14', self.NORMAL_IO_CODE)
upgrade += x64.Jz(self.normal_io_offset - (self.init_driver_offset + len(upgrade.get_code())))
upgrade += x64.Mov('Rax', x64.create_displacement(disp=kldbgdrv + self.HANDLE_ARRAY_ADDR))
upgrade += x64.Label(":LOOP")
upgrade += x64.Mov('RCX', x64.create_displacement('RAX'))
upgrade += x64.Cmp('R14', 'RCX')
upgrade += x64.Jnz(':END')
upgrade += x64.Mov('RAX', x64.create_displacement('RAX', disp=8))
upgrade += x64.Call('RAX')
upgrade += x64.Mov('RBX', 'RAX')
upgrade += x64.JmpAt(kldbgdrv + self.normal_end_offset)
upgrade += x64.Label(":END")
upgrade += x64.Cmp('RCX', 0)
upgrade += x64.Jnz(':NEXT')
upgrade += x64.JmpAt(kldbgdrv + self.fail_offset)
upgrade += x64.Label(":NEXT")
upgrade += x64.Add('RAX', 0x10)
upgrade += x64.Jmp(':LOOP')
self.kdbg.write_pfv_memory(kldbgdrv + self.init_driver_offset, str(upgrade.get_code()))
# Write first array dest
self.write_pfv_ptr(kldbgdrv + self.HANDLE_ARRAY_ADDR, kldbgdrv + self.FIRST_ARRAY_ADDR)
self.write_pfv_ptr(kldbgdrv + self.FIRST_ARRAY_ADDR, 0)
self.write_pfv_ptr(kldbgdrv + self.FIRST_ARRAY_ADDR + 8, 0)
# Jump hijack
jump_init_function = x64.Jmp(self.init_driver_offset - (self.hijack_offset))
self.kdbg.write_pfv_memory(kldbgdrv + self.hijack_offset, str(jump_init_function.get_code()))
self.ioctl_array = kldbgdrv + self.FIRST_ARRAY_ADDR
self.ioctl_array_ptr = kldbgdrv + self.HANDLE_ARRAY_ADDR
self.next_code_addr = kldbgdrv + self.init_driver_offset + len(upgrade.get_code())
self.is_upgraded = True
def test_set_thread_context_64(self, proc64):
code = x64.MultipleInstr()
code += x64.Label(":LOOP")
code += x64.Jmp(":LOOP")
data_len = len(code.get_code())
code += x64.Ret()
t = proc64.execute(code.get_code())
time.sleep(0.1)
assert proc64.is_exit == False
t.suspend()
ctx = t.context
ctx.Rip += data_len
ctx.Rax = 0x11223344
t.set_context(ctx)
t.resume()
time.sleep(0.1)
assert t.exit_code == 0x11223344
def sc_64_LoadLibrary(dll_path):
dll = "KERNEL32.DLL\x00".encode("utf-16-le")
api = "LoadLibraryA\x00"
LoadLibrary64_sc = x64.MultipleInstr()
map(LoadLibrary64_sc.__iadd__, [
shellcraft.amd64.pushstr(dll),
x64.Mov("RCX", "RSP"),
shellcraft.amd64.pushstr(api),
x64.Mov("RDX", "RSP"),
x64.Call(":FUNC_GETPROCADDRESS64"),
x64.Mov("R10", "RAX"),
shellcraft.amd64.pushstr(dll_path),
x64.Mov("RCX", "RSP"),
x64.Sub("RSP", 0x30),
x64.And("RSP", -32),
x64.Call("R10"),
x64.Label(":HERE"),
x64.Jmp(":HERE"), # Dirty infinite loop
# x64.Ret(),
windows.native_exec.nativeutils.GetProcAddress64,
])
return LoadLibrary64_sc.get_code()
import windows
import windows.native_exec.simple_x64 as x64
import windows.native_exec.simple_x86 as x86
from windows.generated_def.winstructs import *
StrlenW64 = x64.MultipleInstr()
StrlenW64 += x64.Label(":FUNC_STRLENW64")
StrlenW64 += x64.Push("RCX")
StrlenW64 += x64.Push("RDI")
StrlenW64 += x64.Mov("RDI", "RCX")
StrlenW64 += x64.Xor("RAX", "RAX")
StrlenW64 += x64.Xor("RCX", "RCX")
StrlenW64 += x64.Dec("RCX")
StrlenW64 += x64.Repne + x64.ScasW()
StrlenW64 += x64.Not("RCX")
StrlenW64 += x64.Dec("RCX")
StrlenW64 += x64.Mov("RAX", "RCX")
StrlenW64 += x64.Pop("RDI")
StrlenW64 += x64.Pop("RCX")
StrlenW64 += x64.Ret()
StrlenA64 = x64.MultipleInstr()
StrlenA64 += x64.Label(":FUNC_STRLENA64")
StrlenA64 += x64.Push("RCX")
StrlenA64 += x64.Push("RDI")
StrlenA64 += x64.Mov("RDI", "RCX")
StrlenA64 += x64.Xor("RAX", "RAX")
StrlenA64 += x64.Xor("RCX", "RCX")
StrlenA64 += x64.Dec("RCX")
StrlenA64 += x64.Repne + x64.ScasB()
