Unicorn Engine

Unicorn Engine هو إطار عمل خفيف لمحاكاة المعالج متعدد المعماريات مبني على QEMU. يتيح لك محاكاة كود الآلة عبر x86 و ARM و AArch64 و MIPS و SPARC ومعماريات أخرى مع تحكم دقيق من خلال تعيين الذاكرة والوصول إلى السجلات وخطافات الأدوات.

التثبيت

# Python binding
pip install unicorn

# From source
git clone https://github.com/unicorn-engine/unicorn.git
cd unicorn
mkdir build && cd build
cmake ..
make -j$(nproc)
sudo make install

# Install Python binding from source
cd ../bindings/python
pip install .

# macOS
brew install unicorn
pip install unicorn

Python API: Basic Emulation

x86-64 Emulation

from unicorn import *
from unicorn.x86_const import *

# Machine code: inc ecx; dec edx; add ecx, edx; ret
X86_CODE = b"\xff\xc1\xff\xca\x01\xd1\xc3"

# Memory address for code
ADDRESS = 0x1000000

# Initialize emulator for x86-64
mu = Uc(UC_ARCH_X86, UC_MODE_64)

# Map 2MB of memory for code
mu.mem_map(ADDRESS, 2 * 1024 * 1024)

# Write code to memory
mu.mem_write(ADDRESS, X86_CODE)

# Set initial register values
mu.reg_write(UC_X86_REG_ECX, 10)
mu.reg_write(UC_X86_REG_EDX, 5)

# Emulate code
mu.emu_start(ADDRESS, ADDRESS + len(X86_CODE) - 1)  # Stop before ret

# Read results
ecx = mu.reg_read(UC_X86_REG_ECX)
edx = mu.reg_read(UC_X86_REG_EDX)
print(f"ECX = {ecx}, EDX = {edx}")  # ECX = 16, EDX = 4

ARM Emulation

from unicorn import *
from unicorn.arm_const import *

# ARM code: mov r0, #5; mov r1, #10; add r2, r0, r1
ARM_CODE = b"\x05\x00\xa0\xe3\x0a\x10\xa0\xe3\x01\x20\x80\xe0"

mu = Uc(UC_ARCH_ARM, UC_MODE_ARM)
mu.mem_map(0x10000, 2 * 1024 * 1024)
mu.mem_write(0x10000, ARM_CODE)

mu.emu_start(0x10000, 0x10000 + len(ARM_CODE))

r0 = mu.reg_read(UC_ARM_REG_R0)
r1 = mu.reg_read(UC_ARM_REG_R1)
r2 = mu.reg_read(UC_ARM_REG_R2)
print(f"R0={r0}, R1={r1}, R2={r2}")  # R0=5, R1=10, R2=15

البنية Setup

from unicorn import *

# x86 32-bit
mu_x86 = Uc(UC_ARCH_X86, UC_MODE_32)

# x86 64-bit
mu_x64 = Uc(UC_ARCH_X86, UC_MODE_64)

# ARM 32-bit
mu_arm = Uc(UC_ARCH_ARM, UC_MODE_ARM)

# ARM Thumb mode
mu_thumb = Uc(UC_ARCH_ARM, UC_MODE_THUMB)

# AArch64
mu_arm64 = Uc(UC_ARCH_ARM64, UC_MODE_ARM)

# MIPS 32-bit big-endian
mu_mips = Uc(UC_ARCH_MIPS, UC_MODE_MIPS32 + UC_MODE_BIG_ENDIAN)

# MIPS 32-bit little-endian
mu_mipsel = Uc(UC_ARCH_MIPS, UC_MODE_MIPS32 + UC_MODE_LITTLE_ENDIAN)

# RISC-V 64-bit
mu_riscv = Uc(UC_ARCH_RISCV, UC_MODE_RISCV64)

Memory Mapping

from unicorn import *
from unicorn.x86_const import *

mu = Uc(UC_ARCH_X86, UC_MODE_64)

# Map memory regions with permissions
CODE_ADDR = 0x400000
STACK_ADDR = 0x7fff0000
DATA_ADDR = 0x600000

# Code: read + execute
mu.mem_map(CODE_ADDR, 0x1000, UC_PROT_READ | UC_PROT_EXEC)

# Stack: read + write
mu.mem_map(STACK_ADDR, 0x10000, UC_PROT_READ | UC_PROT_WRITE)

# Data: read + write
mu.mem_map(DATA_ADDR, 0x1000, UC_PROT_READ | UC_PROT_WRITE)

# Write data to memory
mu.mem_write(DATA_ADDR, b"Hello, Unicorn!\x00")

# Read data from memory
data = mu.mem_read(DATA_ADDR, 16)
print(data)  # b'Hello, Unicorn!\x00'

# Set up stack pointer
mu.reg_write(UC_X86_REG_RSP, STACK_ADDR + 0x8000)

# Unmap memory when done
mu.mem_unmap(DATA_ADDR, 0x1000)

Register Access

from unicorn import *
from unicorn.x86_const import *

mu = Uc(UC_ARCH_X86, UC_MODE_64)
mu.mem_map(0x1000, 0x1000)

# Write general purpose registers
mu.reg_write(UC_X86_REG_RAX, 0xdeadbeef)
mu.reg_write(UC_X86_REG_RBX, 0xcafebabe)
mu.reg_write(UC_X86_REG_RCX, 100)
mu.reg_write(UC_X86_REG_RDX, 200)
mu.reg_write(UC_X86_REG_RSI, 0x600000)
mu.reg_write(UC_X86_REG_RDI, 0x700000)
mu.reg_write(UC_X86_REG_RSP, 0x7fff8000)
mu.reg_write(UC_X86_REG_RBP, 0x7fff8000)
mu.reg_write(UC_X86_REG_RIP, 0x1000)

# Read registers
rax = mu.reg_read(UC_X86_REG_RAX)
rflags = mu.reg_read(UC_X86_REG_EFLAGS)
print(f"RAX = 0x{rax:016x}")
print(f"RFLAGS = 0x{rflags:08x}")

# Read/write segment registers
mu.reg_write(UC_X86_REG_FS_BASE, 0x600000)
fs_base = mu.reg_read(UC_X86_REG_FS_BASE)

Hooks

Code Hooks (Tracing)

from unicorn import *
from unicorn.x86_const import *
from capstone import *

md = Cs(CS_ARCH_X86, CS_MODE_64)

def hook_code(uc, address, size, user_data):
    """Called before every instruction."""
    code = uc.mem_read(address, size)
    for insn in md.disasm(bytes(code), address):
        print(f">>> 0x{insn.address:x}: {insn.mnemonic} {insn.op_str}")

mu = Uc(UC_ARCH_X86, UC_MODE_64)
mu.mem_map(0x1000, 0x1000)
mu.mem_write(0x1000, b"\x48\x31\xc0\x48\xff\xc0\xc3")  # xor rax,rax; inc rax; ret

# Hook all code execution
mu.hook_add(UC_HOOK_CODE, hook_code)

# Hook code in a specific range only
mu.hook_add(UC_HOOK_CODE, hook_code, begin=0x1000, end=0x1010)

mu.emu_start(0x1000, 0x1000 + 6)

Memory Hooks

def hook_mem_read(uc, access, address, size, value, user_data):
    """Called on every memory read."""
    print(f">>> Memory READ at 0x{address:x}, size={size}")

def hook_mem_write(uc, access, address, size, value, user_data):
    """Called on every memory write."""
    print(f">>> Memory WRITE at 0x{address:x}, size={size}, value=0x{value:x}")

def hook_mem_invalid(uc, access, address, size, value, user_data):
    """Called on invalid memory access."""
    print(f">>> Invalid memory access at 0x{address:x}")
    # Map the memory to allow emulation to continue
    uc.mem_map(address & ~0xfff, 0x1000)
    return True  # Continue emulation

mu.hook_add(UC_HOOK_MEM_READ, hook_mem_read)
mu.hook_add(UC_HOOK_MEM_WRITE, hook_mem_write)
mu.hook_add(UC_HOOK_MEM_READ_UNMAPPED | UC_HOOK_MEM_WRITE_UNMAPPED,
            hook_mem_invalid)

Interrupt / Syscall Hooks

def hook_interrupt(uc, intno, user_data):
    """Called on interrupt/syscall."""
    if intno == 0x80:  # Linux int 0x80
        eax = uc.reg_read(UC_X86_REG_EAX)
        print(f">>> Syscall number: {eax}")
        if eax == 1:  # sys_exit
            uc.emu_stop()

def hook_syscall(uc, user_data):
    """Called on x86-64 syscall instruction."""
    rax = uc.reg_read(UC_X86_REG_RAX)
    rdi = uc.reg_read(UC_X86_REG_RDI)
    rsi = uc.reg_read(UC_X86_REG_RSI)
    print(f">>> syscall: rax={rax}, rdi=0x{rdi:x}, rsi=0x{rsi:x}")

mu.hook_add(UC_HOOK_INTR, hook_interrupt)
mu.hook_add(UC_HOOK_INSN, hook_syscall, arg1=UC_X86_INS_SYSCALL)

Emulation Control

from unicorn import *
from unicorn.x86_const import *

mu = Uc(UC_ARCH_X86, UC_MODE_64)
mu.mem_map(0x1000, 0x1000)
code = b"\x90" * 100 + b"\xc3"  # 100 NOPs + ret
mu.mem_write(0x1000, code)

# Start emulation
mu.emu_start(0x1000, 0x1000 + len(code))

# Start with timeout (microseconds)
mu.emu_start(0x1000, 0x1000 + len(code), timeout=5000000)  # 5 seconds

# Start with instruction count limit
mu.emu_start(0x1000, 0x1000 + len(code), count=50)  # Max 50 instructions

# Stop emulation from within a hook
def hook_stop(uc, address, size, user_data):
    if address == 0x1050:
        uc.emu_stop()

mu.hook_add(UC_HOOK_CODE, hook_stop)

Practical: Emulating a Function

from unicorn import *
from unicorn.x86_const import *

def emulate_function(code, args, base=0x400000, stack=0x7fff0000):
    """Emulate an x86-64 function with arguments and return value."""
    mu = Uc(UC_ARCH_X86, UC_MODE_64)

    # Map code and stack
    mu.mem_map(base, 0x10000)
    mu.mem_map(stack, 0x10000)
    mu.mem_write(base, code)

    # Set up stack with return address
    ret_addr = 0xdeadbeef
    sp = stack + 0x8000
    mu.mem_write(sp, ret_addr.to_bytes(8, "little"))
    mu.reg_write(UC_X86_REG_RSP, sp)

    # x86-64 calling convention: RDI, RSI, RDX, RCX, R8, R9
    arg_regs = [UC_X86_REG_RDI, UC_X86_REG_RSI, UC_X86_REG_RDX,
                UC_X86_REG_RCX, UC_X86_REG_R8, UC_X86_REG_R9]
    for i, arg in enumerate(args[:6]):
        mu.reg_write(arg_regs[i], arg)

    # Emulate until ret
    mu.emu_start(base, ret_addr, timeout=10000000)

    return mu.reg_read(UC_X86_REG_RAX)

# Example: emulate add function
# add(a, b) -> a + b
add_code = b"\x48\x89\xf8\x48\x01\xf0\xc3"  # mov rax,rdi; add rax,rsi; ret
result = emulate_function(add_code, [10, 20])
print(f"Result: {result}")  # Result: 30