HackThisSite - Application Mission 1

$ grep -a -o -E 'Contratulations[^\x00]{0,50}' app1win.exe
Contratulations! The password to this level is '
$ grep -a -o -E 'incorrect serial[^\x00]{0,30}' app1win.exe
incorrect serial number. Please re-enter.

#!/usr/bin/env python3
"""Rebuild run-time-built password strings from a Xojo/REALbasic native exe.

REALbasic never stores the level password as a plain string constant.  It
builds it one character at a time at run time:

    mov  ecx, 73h        ; 's'
    push ecx
    call StringDBCSChr   ; -> "s"
    ; 每个字符重复 mov/push/call 这三步
    call RuntimeAddString ; append to the message

so the password is invisible to `strings`.  This script finds the
`mov reg,imm32 ; push reg ; call rel32` idiom, groups call sites by call
target, and prints the immediate characters in address order for every
target fed printable bytes.  The really long runs are the built strings.

usage: ./rebuild_schr.py <exe> [min_run_len]
"""
import struct
76 collapsed lines
import sys
from collections import defaultdict


def load(path):
    """Return (image_base, sections, overlay_file_offset, overlay_bytes)."""
    data = open(path, 'rb').read()
    e_lfanew = struct.unpack_from('<I', data, 0x3C)[0]
    assert data[e_lfanew:e_lfanew + 4] == b'PE\0\0'
    coff = e_lfanew + 4
    nsec = struct.unpack_from('<H', data, coff + 2)[0]
    opt_size = struct.unpack_from('<H', data, coff + 16)[0]
    opt = coff + 20
    magic = struct.unpack_from('<H', data, opt)[0]
    image_base = (struct.unpack_from('<I', data, opt + 28)[0] if magic == 0x10b
                  else struct.unpack_from('<Q', data, opt + 24)[0])
    sec = opt + opt_size
    sections = []
    end = 0
    for i in range(nsec):
        s = sec + i * 40
        name = data[s:s + 8].rstrip(b'\0').decode('latin1')
        vsize, va, rawsize, rawptr = struct.unpack_from('<IIII', data, s + 8)
        sections.append((name, image_base + va, vsize, rawptr, rawsize))
        end = max(end, rawptr + rawsize)
    # REALbasic appends its own object code + string table as a file overlay
    return image_base, sections, end, data[end:]


def char_chains(overlay, base_va, gap=0x120):
    # one chain element is ~0x9c bytes: mov imm32 + push + call SChr + setup +
    # call RuntimeAddString, so the SChr sites of one string sit ~150 bytes apart
    """Find `mov r32,imm ; push r32 ; call rel32` sites and chain their chars."""
    by_target = defaultdict(list)
    i = 0
    while True:
        i = overlay.find(b'\xe8', i)                 # call rel32
        if i < 0 or i + 5 > len(overlay):
            break
        if i >= 2 and 0x50 <= overlay[i - 1] <= 0x57:      # push eax..edi
            r = overlay[i - 1] - 0x50
            if i >= 7 and overlay[i - 6] == 0xB8 + r:      # mov reg, imm32
                val = struct.unpack_from('<I', overlay, i - 5)[0]
                if val <= 0xff:
                    rel = struct.unpack_from('<i', overlay, i + 1)[0]
                    by_target[i + 5 + rel].append((i - 6, val))
        i += 1

    # per-call-site thunks: every SChr call goes through its own stub, so the
    # chain must be rebuilt from *all* char sites in address order, not grouped
    # by call target.
    all_sites = sorted(s for sites in by_target.values() for s in sites)
    runs = []
    cur, prev = [], None
    for off, val in all_sites:
        if prev is not None and off - prev > gap:
            if cur:
                runs.append((0, cur))
            cur = []
        cur.append((off, val))
        prev = off
    if cur:
        runs.append((0, cur))
    return runs


def main():
    path = sys.argv[1]
    min_len = int(sys.argv[2]) if len(sys.argv) > 2 else 8
    image_base, sections, ov_off, overlay = load(path)
    base_va = 0x55E000                       # where REALbasic maps its overlay
    print(f'[*] {path}')
    print(f'[*] overlay at file 0x{ov_off:x}, {len(overlay)} bytes, mapped at 0x{base_va:x}')
    for target, run in sorted(char_chains(overlay, base_va), key=lambda r: -len(r[1])):
        if len(run) < min_len:
            continue
        s = ''.join(chr(v) for _, v in run)
        print(f'\n=== run of {len(run)} chars  ->  {s!r} ===')
        for off, val in run:
            print(f'    file {ov_off + off:#08x}  overlay+{off:#06x}  mov reg, {val:#04x}  {chr(val)!r}')


if __name__ == '__main__':
    main()

$ cd <hts-workspace> && uv run python challenges/hts-app/rebuild_schr.py challenges/hts-app/app1/win/app1win.exe
[*] challenges/hts-app/app1/win/app1win.exe
[*] overlay at file 0x153800, 84840 bytes, mapped at 0x55e000

=== run of 13 chars  ->  'smashthestate' ===
    file 0x15a7f4  overlay+0x6ff4  mov reg, 0x73  's'
    file 0x15a88b  overlay+0x708b  mov reg, 0x6d  'm'
    file 0x15a923  overlay+0x7123  mov reg, 0x61  'a'
    file 0x15a9bb  overlay+0x71bb  mov reg, 0x73  's'
    file 0x15aa53  overlay+0x7253  mov reg, 0x68  'h'
    file 0x15aaeb  overlay+0x72eb  mov reg, 0x74  't'
    file 0x15ab83  overlay+0x7383  mov reg, 0x68  'h'
    file 0x15ac1b  overlay+0x741b  mov reg, 0x65  'e'
    file 0x15acb3  overlay+0x74b3  mov reg, 0x73  's'
    file 0x15ad4b  overlay+0x754b  mov reg, 0x74  't'
    file 0x15ade3  overlay+0x75e3  mov reg, 0x61  'a'
    file 0x15ae7b  overlay+0x767b  mov reg, 0x74  't'
    file 0x15af13  overlay+0x7713  mov reg, 0x65  'e'

$ grep -a -o -E '[0-9]{4}-[0-9]{4}-[0-9]{4}-[0-9]{4}' app1win.exe | sort -u
0130-1414-5624-1341
0138-4411-5902-2411
3810-1941-5861-5351
7692-3349-1914-4567
8361-9811-5511-8134
9484-2341-5696-5321