README.md

md5 hash calculator

Source

https://pwnable.kr/play.php

md5 calculator - under Rookiss

Explanation

As usual, I began by trying to force the system, entering a very long input to see what would happen. Immediately, you can notice a segmentation fault with an indication of smashing the canary. This provides two obvious hints: firstly, there is an option for a stack overflow, and secondly, there is a way to leak the canary.

Let's review the breadcrumbs:

• Using a random number generator with the current time is often a good indication of something insecure.

  __seed = time((time_t *)0x0);
  srand(__seed);

  I Used the rand seed from libc to determine the next rand numbers.

• Using a canary as part of the hash calculation.

  int my_hash(void)
  
  {
  int rand_var;
  int in_GS_OFFSET;
  int i;
  int buff [4];
  int calc3;
  int calc1;
  int calc4;
  int calc2;
  int cannary;
  
  cannary = *(int *)(in_GS_OFFSET + 0x14);
  for (i = 0; i < 8; i = i + 1) {
      rand_var = rand();
      buff[i] = rand_var;
  }
  if (cannary != *(int *)(in_GS_OFFSET + 0x14)) {
      __stack_chk_fail();
  }
  return calc1 + buff[1] + (buff[2] - buff[3]) + cannary + calc2 + (calc3 - calc4);
  }

• Using the system command for something simple.

  system("echo `date` >> log");

Eventually, this program compiled without NX, so there's no ASLR to worry about. The calculations are pretty straightforward this way.

gef➤  info address g_buf
Symbol "g_buf" is at 0x804b0e0 in a file compiled without debugging.

The program decodes (base64) the provided buffer and copies the global buffer, which is much larger, to a smaller buffer on the stack.

From here, I think reading the Python exploit will be much easier.

Thanks for reading.