Tactical LogHuntress CTF 2026 // team archive

Days / Day 05 / Metadata Forensics / Chunk Norris

DAY 05//LANE: METADATA FORENSICS//STATUS: CORRECT x3

Chunk Norris

SSTTMM Modes reported CORRECT in this write-up

Challenge

  • Category: Metadata Forensics (Day 05), GM id metadata-forensics-day-05-chunk-norris
  • Prompt: "The PNGs render, but rendering is not proof. Walk their chunks, verify their declared geometry, and locate the image with an unexpected structural passenger."
  • Question shape (Single Solve): Determine png.ihdr_dimensions_pixels for panel-001.png. (answer_format.kind = exact_text, normalization trim)
  • Flag format: derived dimension string, e.g. WxH

Artifacts

  • metadata-day-05-evidence.zip (challenge artifact, not published) (AES, password infected - same standard password used across this project's artifact zips)
  • Contains panel-001.png (414 bytes)

Triage

Same family as Day 04's "Picture This" (JPEG SOF dimensions), but for PNG: the prompt explicitly distrusts rendering/declared metadata and asks to verify geometry from the file's own structure. For PNG the ground truth lives in the IHDR chunk, which is always the first chunk and always carries the authoritative pixel width/height - independent of anything a viewer or other chunk might claim.

Investigation

  1. Extract the zip with password infected -> panel-001.png.
  2. Parse the PNG structure directly: 8-byte PNG signature, followed by the first chunk. Bytes 8-11 are the IHDR chunk's 4-byte length, bytes 12-15 are the chunk type (IHDR), and bytes 16-23 are two big-endian 32-bit integers: width then height.
  3. Verified directly from the raw bytes:
    magic  = 89 50 4e 47 0d 0a 1a 0a   (PNG signature)
    offset 12..16 = "IHDR"
    width  = struct.unpack('>I', data[16:20])[0] = 269
    height = struct.unpack('>I', data[20:24])[0] = 153

Flag Recovery

png.ihdr_dimensions_pixels for panel-001.png = 269x153, read directly from the IHDR chunk's big-endian width/height fields at byte offsets 16-24. This is a direct structural read, not an inference - no alternate claim (EXIF, filename, or rendered size) was present to compare against in this Single Solve artifact; the full "unexpected structural passenger" outlier (the image whose IHDR disagrees with another declared source) would be found by running the same IHDR parse across the full Mastery-mode multi-panel set and comparing against each image's other metadata, which this single artifact does not include.

Flag

<REDACTED_TOKEN>

(Single Solve, Time Trial, and Mastery were all confirmed CORRECT by the Game Master on 2026-10-05; the raw completion tokens were not preserved locally at solve time, hence the redacted placeholder above - this writeup's technical derivation is independently re-verified from the actual artifact bytes, not reconstructed from memory.)

Minimal Reproduction

import struct, zipfile
z = zipfile.ZipFile('metadata-day-05-evidence.zip')
z.extractall(pwd=b'infected')
data = open('panel-001.png', 'rb').read()
assert data[12:16] == b'IHDR'
w, h = struct.unpack('>II', data[16:24])
print(f'{w}x{h}')  # -> 269x153

Lessons Learned

  • PNG's IHDR chunk is always the first chunk right after the 8-byte signature - no need to walk the full chunk chain just to get dimensions; byte offsets 16-24 are fixed.
  • Same pattern as Day 04's JPEG SOF-dimensions challenge: trust the file's own structural header over any filename, EXIF, or rendering-implied size.
  • Process gap worth fixing: the solving agent did not persist its solver script or state.json for this challenge before closing out, unlike most other Day 05 challenges - this writeup was reconstructed after the fact from the staged artifact and a one-line KB note, and the real completion tokens could not be recovered. Future agents should always save a solver script and state.json immediately after each CORRECT, per the standing project rule, even for a single-field challenge that feels trivial.