How are strings encoded in an ELF file?

Question

I wanted to demonstrate that passwords in clear are easy to read from a program:

#include <stdio.h>
#include <string.h>

int main(int argc, char** argv)
{
    char password[] = "a big refreshing lemonade";
    return strcmp(argv[1], password);
}

But it does not work as expected:

$ gcc foo.c
$ hexdump -C a.out | grep -C2 'lem'
000006c0  00 00 00 48 89 45 f8 31  c0 48 b8 61 20 62 69 67  |...H.E.1.H.a big|
000006d0  20 72 65 48 ba 66 72 65  73 68 69 6e 67 48 89 45  | reH.freshingH.E|
000006e0  d0 48 89 55 d8 48 b8 20  6c 65 6d 6f 6e 61 64 48  |.H.U.H. lemonadH|
000006f0  89 45 e0 66 c7 45 e8 65  00 48 8b 45 c0 48 83 c0  |.E.f.E.e.H.E.H..|
00000700  08 48 8b 00 48 8d 55 d0  48 89 d6 48 89 c7 e8 6d  |.H..H.U.H..H...m|

I notice some weird characters. Why is that?

You likely don't see your explicit string because your code doesn't actually refer to a string literal. Your `char password[]` is a local variable of `main()` that's initialized with the value `"a big refreshing lemonade"`. That doesn't mean the string literal `"a big refreshing lemonade"` has to exist in the executable file - all that has to happen is your array gets properly initialized. Change `char password[]` to `char *password` and you may very well see something different. — Andrew Henle, Jul 19 '19 at 15:29
@AndrewHenle or then not. Nowox: [as if rule.](https://stackoverflow.com/a/46455917/918959) — Antti Haapala -- Слава Україні, Jul 19 '19 at 15:42

score 8 · Accepted Answer · edited Jul 19 '19 at 15:57

It's because the strings aren't being stored as static data.

For example if you had this:

const char* password = "a big refreshing lemonade";

Or even this:

static char password[] = "a big refreshing lemonade";

It is stored contiguously in the binary (You see "a big refreshing lemonade" next to each other) in the constants section.

If you look at the assembly output, you see this:

 6:test.c        ****     char password[] = "a big refreshing lemonade";
23                            .loc 1 6 0
24 001e 48B86120              movabsq $7309940773697495137, %rax
24      62696720
24      7265
25 0028 48BA6672              movabsq $7453010330678293094, %rdx
25      65736869
25      6E67
26 0032 488945D0              movq    %rax, -48(%rbp)
27 0036 488955D8              movq    %rdx, -40(%rbp)
28 003a 48B8206C              movabsq $7233183901389515808, %rax
28      656D6F6E
28      6164
29 0044 488945E0              movq    %rax, -32(%rbp)
30 0048 66C745E8              movw    $101, -24(%rbp)
30      6500

Where you see a lot of movabsq, which loads a 64 bit constant. So, what it does load 8 bytes at a time into password.

You'll notice that the first constant (7309940773697495137) is the little-endian form of "a big re"

score 0 · Answer 2 · answered Jul 19 '19 at 15:50

I wanted to demonstrate that passwords in clear are easy to read from a program...

One can easily use strings, to find this information.

strings - print the strings of printable characters in files.

Per your example,

strings a.out
[ . . . TRUNCATED . . . ]
a big reH
freshingH
 lemonadH
[ . . . TRUNCATED . . . ]

Full output,

strings a.out
/lib64/ld-linux-x86-64.so.2
libc.so.6
__stack_chk_fail
__cxa_finalize
strcmp
__libc_start_main
GLIBC_2.2.5
GLIBC_2.4
_ITM_deregisterTMCloneTable
__gmon_start__
_ITM_registerTMCloneTable
a big reH
freshingH
 lemonadH
AWAVI
AUATL
[]A\A]A^A_
;*3$"
GCC: (Ubuntu 7.4.0-1ubuntu1~18.04.1) 7.4.0
crtstuff.c
deregister_tm_clones
__do_global_dtors_aux
completed.7697
__do_global_dtors_aux_fini_array_entry
frame_dummy
__frame_dummy_init_array_entry
foo.c
__FRAME_END__
__init_array_end
_DYNAMIC
__init_array_start
__GNU_EH_FRAME_HDR
_GLOBAL_OFFSET_TABLE_
__libc_csu_fini
_ITM_deregisterTMCloneTable
_edata
__stack_chk_fail@@GLIBC_2.4
__libc_start_main@@GLIBC_2.2.5
__data_start
strcmp@@GLIBC_2.2.5
__gmon_start__
__dso_handle
_IO_stdin_used
__libc_csu_init
__bss_start
main
__TMC_END__
_ITM_registerTMCloneTable
__cxa_finalize@@GLIBC_2.2.5
.symtab
.strtab
.shstrtab
.interp
.note.ABI-tag
.note.gnu.build-id
.gnu.hash
.dynsym
.dynstr
.gnu.version
.gnu.version_r
.rela.dyn
.rela.plt
.init
.plt.got
.text
.fini
.rodata
.eh_frame_hdr
.eh_frame
.init_array
.fini_array
.dynamic
.data
.bss
.comment
root@42d62eac5ccf:~#
root@42d62eac5ccf:~# strings a.out
/lib64/ld-linux-x86-64.so.2
libc.so.6
__stack_chk_fail
__cxa_finalize
strcmp
__libc_start_main
GLIBC_2.2.5
GLIBC_2.4
_ITM_deregisterTMCloneTable
__gmon_start__
_ITM_registerTMCloneTable
a big reH
freshingH
 lemonadH
AWAVI
AUATL
[]A\A]A^A_
;*3$"
GCC: (Ubuntu 7.4.0-1ubuntu1~18.04.1) 7.4.0
crtstuff.c
deregister_tm_clones
__do_global_dtors_aux
completed.7697
__do_global_dtors_aux_fini_array_entry
frame_dummy
__frame_dummy_init_array_entry
foo.c
__FRAME_END__
__init_array_end
_DYNAMIC
__init_array_start
__GNU_EH_FRAME_HDR
_GLOBAL_OFFSET_TABLE_
__libc_csu_fini
_ITM_deregisterTMCloneTable
_edata
__stack_chk_fail@@GLIBC_2.4
__libc_start_main@@GLIBC_2.2.5
__data_start
strcmp@@GLIBC_2.2.5
__gmon_start__
__dso_handle
_IO_stdin_used
__libc_csu_init
__bss_start
main
__TMC_END__
_ITM_registerTMCloneTable
__cxa_finalize@@GLIBC_2.2.5
.symtab
.strtab
.shstrtab
.interp
.note.ABI-tag
.note.gnu.build-id
.gnu.hash
.dynsym
.dynstr
.gnu.version
.gnu.version_r
.rela.dyn
.rela.plt
.init
.plt.got
.text
.fini
.rodata
.eh_frame_hdr
.eh_frame
.init_array
.fini_array
.dynamic
.data
.bss
.comment

The `strings` should be probably just a comment. The strings are as much readable within the hexdump too. — Antti Haapala -- Слава Україні, Jul 19 '19 at 19:00

How are strings encoded in an ELF file?

2 Answers2