Description

The server lets you upload two PDF files. It grants the flag only if both files have the same MD5 hash but different contents. This requires exploiting a known weakness in the MD5 algorithm.

Overview

• The challenge tests whether two uploaded files produce an identical MD5 hash while having different byte content.
• MD5 is a broken cryptographic hash function: researchers demonstrated practical collision generation in 2004, and pre-computed collision pairs have been publicly available since.
• The solution is to obtain a known MD5 collision pair and submit those two files.

Background: MD5 Collision

• A hash function collision occurs when two distinct inputs produce the same hash output.
• MD5 is particularly weak because collisions can be generated in seconds on modern hardware using the Chosen-Prefix Collision technique.
• The fastcoll tool by Marc Stevens generates pairs of files with identical MD5 hashes given any shared prefix.
• In practice, pre-computed PDF collision pairs are widely available for exactly this type of challenge.

Solution

Step 1: Obtain a Collision Pair

• Two PDFs (hello.pdf and erase.pdf) are prepared such that md5(hello.pdf) == md5(erase.pdf) but their byte contents differ.
• These can be generated with fastcoll or downloaded from known collision repositories.

Step 2: Upload Both Files

• The server expects a multipart/form-data POST with fields file1, file2, and submit.
• requests_toolbelt.MultipartEncoder is used to preserve the correct content type boundary.

from requests_toolbelt import MultipartEncoder

mp_encoder = MultipartEncoder(fields={
    "file1": ("hello.pdf", open("hello.pdf", "rb"), "application/pdf"),
    "file2": ("hello.pdf", open("erase.pdf", "rb"), "application/pdf"),
    "submit": "Upload",
})

r = requests.post(url, data=mp_encoder,
                  headers={"Content-Type": mp_encoder.content_type})

• Both files are uploaded with the filename hello.pdf so the server does not reject them on name mismatch.

Step 3: Extract the Flag

• The server confirms the MD5 match and returns the flag in the HTML response body.

soup = BeautifulSoup(r.text, "html.parser")
result = [c for c in soup.get_text(separator="\n").splitlines()
          if "picoCTF{" in c][0]

Flag

picoCTF{c0ngr4ts_u_r_1nv1t3d_73b0c8ad}