Skip to content

Latest commit

 

History

History
 
 

elrond32

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 

tinyCTF 2014: Elrond32

Category: Reverse Engineering Points: 300 Description:

Download file

Write-up

Let’s extract the provided rev300.zip file:

$ unzip rev300.zip
Archive:  rev300.zip
  inflating: rev300

The extracted rev300 file is an ELF binary:

$ file rev300
rev300: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.24, stripped

We open it in IDA. As usual, the main function is the first argument to ___libc_start_main. Its functionality is simple: if argv[1] is present it calls check_password(argv[1], 0). If this check succeeds it calls print_flag(argv[1]). Since print_flag depends on argv[1] we must first reverse check_password.

Rename the arguments to get check_password(password, pos). At the beginning of the function it uses a jump table based on pos, meaning in the original code a switch statement was used. We have to manually tell IDA this. Go to the jump eax instruction, then EditOtherSpecify Switch Idiom. The jump table has 10 elements, each a pointer of size 4 bytes. The default case is located at 0x80484D5. We end up with a nice layout after this:

Jump table

Inspecting the function now reveals that in each call the first character is checked (in one of the switch cases). If this character is correct, it recursively calls check_password(password + 1, newpos). Once pos equals 2 or 9, the default switch case is hit, which returns true (i.e. the password is correct). The final difficulty is finding out how newpos is calculated in the recursive call [1]. A commented version is shown in the following image:

Calculation of newpos

The magic constant 0x2E8BA2E9 is used to perform an optimized division by 11. This optimization technique replaces a constant division by a multiplication of its reciprocal in fixed-point notation. One trick to find out the denominator is to divide 2^32 by the magic constant, and multiply this to match any shift operation like sar edx, 1. In this case we have 2^32 / 0x2E8BA2E9 * 2 which equals 11. The other arithmetic operations are straightforward. So the value of the new pos is (keep in mind that all intermediate results are rounded down):

newpos = (pos + 1) * 7 - [(pos + 1) * 7 / 11 * 11]

This is equivalent with:

temp = (pos + 1) * 7
help = temp - (temp % 11)
newpos = temp - help

Which can be simplified to:

newpos = temp - help
       = temp - temp + (temp % 11)
       = temp % 11
       = [(pos + 1) * 7] % 11

Since the first value of pos is 0, the subsequent values are 7, 1, 3, 6, 5, 9, 4, and finally 2 (which results in a default switch case). Recall that in each of the resulting switch cases, the first character is checked. Hence the password must be isengard. This prints out the flag:

flag{s0me7hing_S0me7hinG_t0lki3n}

[1] Here I decided to explain how to reverse this calculation, since I think this is useful to know. Another approach is to simply step through the execution using a debugger, and dynamically observe which characters are compared to what. This can even be scripted in gdb.

Alternate solution

Using the hint of Elrond, @MajorD4m4ge downloaded a Tolkien dictionary from http://packetstorm.igor.onlinedirect.bg/Crackers/wordlists/literature/tolkien.gz and ran:

$ zcat tolkien.gz | while read line; do ./rev300 "$line"; done | grep -v denied
Access granted
flag{s0me7hing_S0me7hinG_t0lki3n}

Running it on rockyou.txt.gz eventually revealed the flag as well.

Other write-ups and resources