IDA Python 使用总结_idapython，2024年最新网络安全大厂面试笔试题分享

最新推荐文章于 2024-05-03 11:33:07 发布

美好前端开发者

最新推荐文章于 2024-05-03 11:33:07 发布

阅读量775

点赞数 9

分类专栏： 2024年程序员学习文章标签： python web安全面试

本文链接：https://blog.csdn.net/m0_61549674/article/details/137353566

版权

2024年程序员学习专栏收录该内容

257 篇文章 0 订阅

订阅专栏

for i in range(5):
mnem = idc.print_insn_mnem(ea)
assert mnem == mnem_list[i]
if mnem == ‘mov’:
imm = idc.get_operand_value(ea, 1)
reg = idc.print_operand(ea, 0)
elif mnem == ‘call’:
call_target = idc.get_operand_value(ea, 0)
ea += idc.get_item_size(ea)
return Block(start_ea, ea, imm, reg, call_target)

def handle_one_branch(branch_address, new_code_ea):
new_code = b’’
ea = branch_address
while True:
try:
block = get_block(ea)
is_ret, real_code = get_real_code(block, new_code_ea)
reloc.merge(ea, new_code_ea)
ea = block.end_ea
new_code_ea += len(real_code)
new_code += real_code
if is_ret: break
except:
get_eip_func = {0x900: ‘ebx’, 0x435c: ‘eax’}
if idc.print_insn_mnem(ea) == ‘call’ and get_operand_value(ea, 0) in get_eip_func:
reloc.merge(ea, new_code_ea)
real_code = asm(‘mov %s, 0x%x’ % (get_eip_func[get_operand_value(ea, 0)], ea + 5), new_code_ea)
else:
if idc.print_insn_mnem(ea) == ‘jmp’ and idc.get_operand_type(ea, 0) != idc.o_reg:
reloc.merge(new_code_ea, ea)
else:
reloc.merge(ea, new_code_ea)
real_code = mov_code(ea, new_code_ea)

new_code += real_code
if real_code == asm(‘ret’): break
new_code_ea += len(real_code)
if idc.print_insn_mnem(ea) == ‘jmp’ and idc.get_operand_type(ea, 0) != idc.o_reg: # jmp reg is a swtich
jmp_ea = idc.get_operand_value(ea, 0)
if reloc.get(jmp_ea)[1] == False: break # 跳回之前的代码说明是个循环
ea = reloc.get(jmp_ea)[0]
else:
ea += get_item_size(ea)
return new_code

def solve():
entry_point = 0x48F4
new_code_start = 0x96150
new_code_ea = new_code_start

jmp_table = (0x892ac, 0x8c000) # [0x8c000 + (eax>>2) - 0x2d54] + 0x8c000

for _ in range(0x10000): idc.del_items(new_code_ea + _)
ida_bytes.patch_bytes(new_code_ea, 0x10000 * b’\x90’)

func_queue = Queue()
func_queue.put(entry_point)

while not func_queue.empty():
func_address = func_queue.get()
if reloc.get(func_address)[1] == False: continue
reloc.merge(func_address, new_code_ea)
branch_queue = Queue()
branch_queue.put(func_address)
if func_address == 0x4148: # 特判 0x4148 地址处的函数，读取跳转表。
assert new_code_ea == 0x963d0
for eax in range(0x20):
jmp_target = (ida_bytes.get_dword(jmp_table[0] + eax * 4) + jmp_table[1]) & 0xFFFFFFFF
new_jmp_target, need_handle = reloc.get(jmp_target)
if need_handle: branch_queue.put(jmp_target)

while not branch_queue.empty():
branch_address = branch_queue.get()
new_code = handle_one_branch(branch_address, new_code_ea)
ida_bytes.patch_bytes(new_code_ea, new_code)

当前 branch 去完混淆之后需要遍历代码找到 call 和 jmp 指令从而找到其他的 function 和 branch 。

ea = new_code_ea
while ea < new_code_ea + len(new_code):
idc.create_insn(ea)
if idc.print_insn_mnem(ea) == ‘call’:
call_target, need_handle = reloc.get(get_operand_value(ea, 0))
if need_handle: func_queue.put(call_target)
elif idc.print_insn_mnem(ea)[0] == ‘j’ and idc.get_operand_type(ea, 0) != idc.o_reg:
jcc_target, need_handle = reloc.get(get_operand_value(ea, 0))
if need_handle == True:
branch_queue.put(jcc_target)
ea += get_item_size(ea)
new_code_ea += len(new_code)

ea = new_code_start
while ea < new_code_ea:
idc.create_insn(ea)
mnem = idc.print_insn_mnem(ea)

if mnem == ‘call’:
call_target, need_handle = reloc.get(get_operand_value(ea, 0))
assert need_handle == False
ida_bytes.patch_bytes(ea, asm(‘call 0x%x’ % (call_target), ea))
elif mnem[0] == ‘j’ and idc.get_operand_type(ea, 0) != idc.o_reg:
jcc_target, need_handle = reloc.get(get_operand_value(ea, 0))
assert need_handle == False
ida_bytes.patch_bytes(ea, asm(‘%s 0x%x’ % (mnem, jcc_target), ea).ljust(idc.get_item_size(ea), b’\x90’))
elif mnem == ‘pushf’:
ida_bytes.patch_bytes(ea, b’\x90’ * 9)
ea += 9
continue
ea += get_item_size(ea)

new_jmp_table = (0xA6000 - 0x2D54, 0xA6000)

移动并修复跳转表

for eax in range(0x20):
jmp_target = (ida_bytes.get_dword(jmp_table[0] + eax * 4) + jmp_table[1]) & 0xFFFFFFFF
new_jmp_target, need_handle = reloc.get(jmp_target)
assert need_handle == False
ida_bytes.patch_dword(new_jmp_table[0] + eax * 4, (new_jmp_target - new_jmp_table[1]) & 0xFFFFFFFF)

need_patch_addr = 0x963D7
ida_bytes.patch_bytes(need_patch_addr, asm(‘call 0x900;add ebx, 0x%x’ % (new_jmp_table[1] - (need_patch_addr + 5)), need_patch_addr)) # 修复指令
ida_bytes.patch_bytes(new_jmp_table[1] - 0x2d7a, ida_bytes.get_bytes(jmp_table[1] - 0x2d7a, 0x26)) # 复制字符串到正确位置

for _ in range(0x10000): idc.del_items(new_code_ea + _)
idc.jumpto(new_code_start)
ida_funcs.add_func(new_code_start)

print(“finish”)

solve()

例题：SUSCTF2022 tttree

附件下载链接

首先将 0x140010074 ，0x140017EFA ，140018C67 起始处的数据转换为汇编。

观察汇编，发现很多代码块之间相互跳转，因此先按照 retn 划分代码块。通过对代码块的观察，发现这些代码块按照 call $+5;pop rax（即 E8 00 00 00 00 58 ）的出现次数可以分为三种：

出现 0 次：

本质上是 其它操作 + retn 。
出现 1 次：

这种代码块本质为 其它操作 + jmp target ，注意 其它操作 中可能包含 branch 。
出现 2 次：

这个可以看做 2 个出现 1 次的代码块两个拼在一起，其中前面一个代码块去掉 retn 。执行完前面一个代码块后由于没有 retn ，因此 target1 留在栈中。执行第 2 个代码块跳转到 target2 执行，在 target2 代码块返回时会返回到 target1 。因此这种代码块本质上相当于 其它操作 + call target2 且下一个要执行的代码块为 target1 。

我们定义代码块 Block 几个关键信息：

start_addr：代码块的起始地址。
asm_list：代码块的有效汇编，由于汇编指令可能包含 [rip + xxx] ，因此需要记录汇编指令的地址以便后续修正。
direct_next：执行完此代码块后接下来要执行的代码块地址。
branch_list：代码块中的所有条件跳转语句跳到的地址。
call_target：代码块调用函数地址。

class Block:
def __init__(self, start_ea, asm_list, direct_next, branch_list, call_target):
self.start_ea = start_ea
self.asm_list = asm_list
self.direct_next = direct_next
self.branch_list = branch_list
self.call_target = call_target

def __str__(self):
return ‘start_ea: 0x%x\ndirect_next: 0x%x\ncall_target: 0x%x\nbranch_list: %s\nasm_list:\n%s\n’ % (
0 if self.start_ea == None else self.start_ea,
0 if self.direct_next == None else self.direct_next,
0 if self.call_target == None else self.call_target,
str([hex(x) for x in self.branch_list]),
str(‘\n’.join([hex(addr) + ’ ’ + asm for addr, asm in self.asm_list]))
)

get_block 函数可以获取给定地址处的代码块并提取相关信息。代码块中可能有 push xxx;pop xxx; 这样的无意义指令，可以通过栈模拟来去除。

def get_block(start_ea):
ea = start_ea
stack = []
asm_list = []
branch_list = []
call_target = None
direct_next = None

while True:
idc.create_insn(ea)
mnem = idc.print_insn_mnem(ea)

处理混淆中跳转的情况

if mnem == ‘pushfq’:
ea += idc.get_item_size(ea)

assert idc.get_bytes(ea, idc.get_item_size(ea)) == b’\xE8\x00\x00\x00\x00’
ea += idc.get_item_size(ea)
jmp_base = ea

assert idc.print_insn_mnem(ea) == ‘pop’ and idc.get_operand_type(ea, 0) == o_reg
reg = idc.print_operand(ea, 0)
ea += idc.get_item_size(ea)

assert idc.print_insn_mnem(ea) == ‘add’ and idc.print_operand(ea, 0) == reg
assert idc.get_operand_type(ea, 1) == o_imm

jmp_target = (jmp_base + idc.get_operand_value(ea, 1)) & 0xFFFFFFFFFFFFFFFF
ea += idc.get_item_size(ea)

assert idc.get_bytes(ea, idc.get_item_size(ea)) == asm(‘mov [rsp + 0x10], %s’ % reg, ea)
ea += idc.get_item_size(ea)

assert idc.print_insn_mnem(ea) == ‘popfq’
ea += idc.get_item_size(ea)

assert idc.print_insn_mnem(ea) == ‘pop’ and idc.print_operand(ea, 0) == reg
assert len(stack) != 0 and stack[-1][0] == ‘push’ and stack[-1][1] == reg
stack.pop()
asm_list.pop()

assert len(stack) != 0 and stack[-1][0] == ‘push’ and stack[-1][1] == reg
stack.pop()
asm_list.pop()

ea += idc.get_item_size(ea)

if idc.print_insn_mnem(ea) == ‘retn’:
if direct_next == None:
direct_next = jmp_target
elif call_target == None:
call_target = jmp_target
asm_list.append((0, ‘call 0x%x’ % (call_target)))
else:
print(“🤔🤔🤔🤔🤔🤔”)
assert False
break
else:
assert call_target == None and direct_next == None
direct_next = jmp_target
continue

if mnem == ‘push’:
stack.append((mnem, idc.print_operand(ea, 0)))
elif mnem == ‘pop’:
if len(stack) != 0 and stack[-1][0] == ‘push’ and stack[-1][1] == idc.print_operand(ea, 0):
stack.pop()
asm_list.pop()
ea += idc.get_item_size(ea)
continue
else:
stack.clear()
else:
stack.clear()

asm_list.append((ea, disasm(idc.get_bytes(ea, idc.get_item_size(ea)), ea)))

if mnem == ‘retn’: break
if mnem[0] == ‘j’ and mnem != ‘jmp’ and idc.get_operand_type(ea, 0) != o_reg:
branch_list.append(idc.get_operand_value(ea, 0))

if mnem == ‘jmp’:
if idc.get_segm_name(idc.get_operand_value(ea, 0)) not in [‘.text’, ‘.aaa’]:
break
else:
ea = idc.get_operand_value(ea, 0)
else:
ea += idc.get_item_size(ea)

return Block(start_ea, asm_list, direct_next, branch_list, call_target)

能够获取代码块信息之后就可以 bfs 函数以及函数中的所有分支，提取出汇编代码并写入 newcode 段。这里需要注意以下几点：

涉及 rip 的汇编指令不能只是简单把指令中的 rip 替换为对应的具体数值，因为有的指令立即数的长度被限制在 4 字节，直接替换成数值会溢出。一个比较好的解决方法是将 rip 替换为 rip + (指令原本地址 - 指令当前地址) 。这样借助 rip 寄存器扩大访问范围并且代码移动的距离不会超过 0x100000000 因此可以保证正确性。
如果 block.direct_next 对应的代码已经被去混淆了需要加上一条 jmp 指令跳转到已经去混淆的代码。
有的汇编指令 keystone 不支持汇编，比如 bnd ret ，需要特判。

while not func_queue.empty():
func_address = func_queue.get()
if reloc.get(func_address)[1] == False: continue
branch_queue = Queue()
branch_queue.put(func_address)

while not branch_queue.empty():
branch_address = branch_queue.get()
ea = branch_address

while True:
block = get_block(ea)
reloc.merge(ea, new_code_ea)

for addr, insn in block.asm_list:
insn = insn.replace(‘rip’, ‘rip - 0x%x’ % (new_code_ea - addr))
if insn == ‘bnd ret ;’:
code = b’\xF2\xC3’
else:
code = asm(insn, new_code_ea)
ida_bytes.patch_bytes(new_code_ea, code)
if addr != 0: reloc.merge(addr, new_code_ea)
new_code_ea += len(code)

if block.call_target != None:
call_target, need_handle = reloc.get(block.call_target)
if need_handle: func_queue.put(call_target)

for branch_address in block.branch_list:
jcc_target, need_handle = reloc.get(branch_address)
if need_handle: branch_queue.put(jcc_target)

if block.direct_next == None: break

next_target, need_handle = reloc.get(block.direct_next)
if need_handle == False:
code = asm(‘jmp 0x%x’ % (next_target), new_code_ea)
ida_bytes.patch_bytes(new_code_ea, code)
new_code_ea += len(code)
break
else:
ea = block.direct_next

最后对代码进行重定位，需要注意的是代码块中的有效指令中也可能有 call 指令，这里 call 调用的是一个类似 plt 表的结构，会直接跳转到导入表中的函数地址表指向的函数，需要特判这种情况。

ea = new_code_start
while ea < new_code_ea:
assert idc.create_insn(ea) != 0
mnem = idc.print_insn_mnem(ea)

if mnem == ‘call’:
call_target, need_handle = reloc.get(get_operand_value(ea, 0))
if need_handle == True:
if idc.print_insn_mnem(call_target) == ‘jmp’ and idc.get_segm_name(idc.get_operand_value(call_target, 0)) == ‘.idata’:
ea += get_item_size(ea)
continue
else:
assert False
ida_bytes.patch_bytes(ea, asm(‘call 0x%x’ % (call_target), ea).ljust(idc.get_item_size(ea), b’\x90’))
elif mnem[0] == ‘j’ and idc.get_operand_type(ea, 0) != idc.o_reg:
jcc_target, need_handle = reloc.get(get_operand_value(ea, 0))
assert need_handle == False
ida_bytes.patch_bytes(ea, asm(‘%s 0x%x’ % (mnem, jcc_target), ea).ljust(idc.get_item_size(ea), b’\x90’))

ea += get_item_size(ea)

最后完整代码：

from queue import *
from idc import *
import idc
from keystone import *
from capstone import *

asmer = Ks(KS_ARCH_X86, KS_MODE_64)
disasmer = Cs(CS_ARCH_X86, CS_MODE_64)

def disasm(machine_code, addr=0):
l = “”
for i in disasmer.disasm(machine_code, addr):
l += “{:8s} {};\n”.format(i.mnemonic, i.op_str)
return l.strip(‘\n’)

def asm(asm_code, addr=0):
l = b’’
for i in asmer.asm(asm_code, addr)[0]:
l += bytes([i])
return l

class RelocDSU:

def __init__(self):
self.reloc = {}

def get(self, ea):
if ea not in self.reloc:
if idc.print_insn_mnem(ea) == ‘jmp’ and idc.get_operand_type(ea, 0) != idc.o_reg:
jmp_ea = idc.get_operand_value(ea, 0)

if idc.get_segm_name(jmp_ea) == ‘.idata’:
self.reloc[ea] = ea
return self.reloc[ea], False

self.reloc[ea], need_handle = self.get(idc.get_operand_value(ea, 0))
return self.reloc[ea], need_handle
else:
self.reloc[ea] = ea
if self.reloc[ea] != ea: self.reloc[ea] = self.get(self.reloc[ea])[0]
return self.reloc[ea], idc.get_segm_name(self.reloc[ea]) in [‘.text’, ‘.aaa’]

def merge(self, ea, reloc_ea):

print((hex(ea), hex(reloc_ea)))

self.reloc[self.get(ea)[0]] = self.get(reloc_ea)[0]

reloc = RelocDSU()