CNIT 127 Ch 3: Shellcode

CNIT 127: Exploit Development 
 
Ch 3: Shellcode
Updated 2-2-18

Topics
• Protection rings
• Syscalls
• Shellcode
• nasm Assembler
• ld GNU Linker
• objdump to see contents of object files
• strace System Call Tracer
• Removing Nulls
• Spawning a Shell

Shellcode
• Written in assembler
• Translated into hexadecimal opcodes
• Intended to inject into a system by
exploiting a vulnerability
• Typically spawns a root shell, but may do
something else

System Calls (or Syscalls)
• Syscalls directly access the kernel, to:
– Get input
– Produce output
– Exit a process
– Execute a binary file
– And more
• They are the interface between protected
kernel mode and user mode

Protection Rings
• Although the x86
provides four
rings, only rings 0
and 3 are used by
Windows or Unix
• Ring 3 is user-
land
• Ring 0 is kernel-
land
• Links Ch 3a-3c

Protecting the Kernel
• Protected kernel mode
– Prevents user applications from compromising
the OS
• If a user mode program attempts to
access kernel memory, this generates an
access exception
• Syscalls are the interface between user
mode and kernel mode

Libc
• C library wrapper
• C functions that perform syscalls
• Advantages of libc
– Allows programs to continue to function
normally even if a syscall is changed
– Provides useful functions, like malloc
– (malloc allocates space on the heap)
• See link Ch 3d

Syscalls use INT 0x80
1. Load syscall number into EAX
2. Put arguments in other registers
3. Execute INT 0x80
4. CPU switches to kernel mode
5. Syscall function executes

Syscall Number and Arguments
• Syscall number is an integer in EAX
• Up to six arguments are loaded into
– EBX, ECX, EDX, ESI, EDI, and EBP
• For more than six arguments, the first
argument holds a pointer to a data
structure

exit()
• The libc exit function does a lot of
preparation, carefully covering many
possible situations, and then calls SYSCALL
to exit

Disassembling exit
• gdb e
– disassemble main
– disassemble exit
– disassemble __run_exit_handlers
• All that stuff is error handling, to prepare
for the syscall, which is at the label _exit
• disassemble _exit

Disassembling _exit
• syscall 252 (0xfc), exit_group() (kill all threads)
• syscall 1, exit() (kill calling thread)
– Link Ch 3e

Writing Shellcode for the
exit() Syscall

Shellcode Size
• Shellcode should be as simple and
compact as possible
• Because vulnerabilities often only allow a
small number of injected bytes
– It therefore lacks error-handling, and will
crash easily

Seven Instructions
• exit_group
• exit

sys_exit Syscall
• Two arguments: eax=1, ebx is return value
(0 in our case)
• Link Ch 3m

nasm and ld
• nasm creates object file
• ld links it, creating an executable ELF file

objdump
• Shows the contents of object files

C Code to Test Shellcode
• From link Ch 3k
• Textbook version explained at link Ch 3i

Compile and Run
• Textbook omits the "-z execstack" option
• It's required now or you get a segfault
• Next, we'll use "strace" to see all system
calls when this program runs
• That shows a lot of complex calls, and
"exit(0)" at the end

Using strace
• apt-get install strace

Getting Rid of Nulls
• We have null bytes, which will terminate
a string and break the exploit

Replacing Instructions
• This instruction contains nulls
– mov ebx,0
• This one doesn't
– xor ebx,ebx
• This instruction contains nulls, because it
moves 32 bits
– mov eax,1
• This one doesn't, moving only 8 bits
– mov al, 1

Beyond exit()
• The exit() shellcode stops the program, so
it just a DoS attack
• Any illegal instruction can make the
program crash, so that's of no use
• We want shellcode that offers the
attacker a shell, so the attacker can type
in arbitrary commands

Five Steps to Shellcode
1. Write high-level code
2. Compile and disassemble
3. Analyze the assembly
4. Clean up assembly, remove nulls
5. Extract commands and create shellcode

fork() and execve()
• Two ways to create a new process in Linux
• Replace a running process
– Uses execve()
• Copy a running process to create a new
one
– Uses fork() and execve() together

C Program to Use execve()
• See link Ch 3l

Recompile with Static
• Static linking preserves our execve syscall
• objdump of main is long, but we only care
about main and __execve

main()
• Pushes 3 Arguments
• Calls __execve

Man Page
• execve() takes three arguments

execve() Arguments
1. Pointer to a string containing the name
of the program to execute
– "/bin/sh"
2. Pointer to argument array
– happy
3. Pointer to environment array
– NULL

Objdump of __execve
• Puts four parameters into edx, ecx, ebx,
and eax
• INT 80

CNIT 127 Ch 3: Shellcode

More Related Content

CNIT 127 Ch 3: Shellcode