Tactical LogHuntress CTF 2026 // team archive

Days / Day 28 / Binary RE / Family Resemblance

DAY 28//LANE: BINARY RE//STATUS: CORRECT x3

Family Resemblance

Binary RE SSTTMM Modes reported CORRECT in this write-up

Family Resemblance — solution report

Outcome

All three requested modes are complete. The Game Master status endpoint reports CORRECT for each attempt.

ModeAccepted fieldsServer resultFlag
Single Solve392 / 392CORRECTTokenfe52b075d936b1229339c21908d38441
Time Trial392 / 392CORRECTToken110b360b0c9f42b53fd0e6195b9f2794
Mastery4,704 / 4,704CORRECTTokeneea515f5b8fede9be1c23945c6938e6c

The raw start responses, submit responses, and final status responses are kept in this directory and its mode subdirectories. The full player-facing questions, exact answer IDs, formats, artifact links, hashes, and timer fields are rendered in CHALLENGE.md, the Time Trial brief, and the Mastery brief.

Family classification

For the Single Solve Day 28 specimens, family-a.elf (challenge artifact, not published) and family-b.elf (challenge artifact, not published) are the matching family. false-friend.elf (challenge artifact, not published) shares the same broad executable appearance but is the stranger.

SpecimenSizeSHA-256ELF / machine.text SHA-256HCFG campaign
family-a.elf (challenge artifact, not published)22,922374a67f58a81bb0d223cacfbeac6062a2b79d34ff088eab3d46353cb464b3716ELF64, x86-64d1bc8c5d4625420d3a143a9433f39cdc160678f732d723952cbee5ee015265b9BR28-000
family-b.elf (challenge artifact, not published)22,922b79f32a01f0177121fe0fdeca8b9dd4ab32630427c7e2031785285bb8d237cdeELF64, x86-64d1bc8c5d4625420d3a143a9433f39cdc160678f732d723952cbee5ee015265b9BR28-000
false-friend.elf (challenge artifact, not published)22,923f50f37d7bd263706d06648d46774a62b392d588c4bcd09733f5ee432868764beELF64, x86-64df62f7ec83188a16931c0c9a559ee57651c4b3a89494031cffa538156bc636d0FF28-000

All three have the same ELF class, little-endian encoding, ET_DYN type, entry point 0x10e0, machine value 62, and section-name sequence. That metadata alone makes the false friend convincing.

The two family specimens have identical .text bytes and their decoded HCFG JSON objects are byte-for-byte equal. Their whole files differ at only one byte (offset 12,824). The false friend has a different .text hash and a separate config:

  • Family config: domain node-000.day28.example.invalid, IPv4 198.51.100.50, mutex Global\HuntressBR28_000, URL https://node-000.day28.example.invalid/api/BR28-000, service HuntressLab-Day28-Family-000.
  • False-friend config: domain decoy-000.day28.example.invalid, IPv4 198.51.100.70, mutex Global\HuntressFF28_000, URL https://decoy-000.day28.example.invalid/api/FF28-000, service HuntressLab-Day28-False-000.

This is why header and filename similarity were treated as weak evidence. Matching behavior and matching decoded configuration identified the family; the changed code section and decoy configuration identified the false friend.

Analysis method

  1. Started each mode through the Game Master proxy and saved the complete start response. Time Trial and Mastery used run_for_time: true.
  2. Downloaded each assigned evidence.zip (challenge artifact, not published) through ?op=artifact. For Mastery, downloaded all 12 deliveries by delivery index and checked every SHA-256 against the start response before analysis.
  3. Inspected the ZIP central directory and preserved its member order. The bundles use WinZip AES encryption; pyzipper extracted the exact bytes with the lab password infected.
  4. Parsed each format from its bytes:
    • ELF headers, sections, symbol tables, section data, and disassembly. ELF symbol ordering is retained for questions that require symbol-table order.
    • PE/COFF headers, section tables, machine codes, and primary COFF symbols. Long COFF names are resolved through the string table; auxiliary records are skipped.
    • CLR metadata-root and CLI runtime fields for managed PE files.
    • JAR central-directory entries, manifest Main-Class, Java classfile constant pools, method tables, and access flags.
    • WebAssembly magic and section structure where requested.
  5. Decoded HCFG records only after validating them. A record is HCFG, version byte, XOR key byte, little-endian payload length, little-endian CRC32, then XOR-encoded JSON. Base64 carriers were decoded and validated the same way; invalid lookalike records were ignored.
  6. Decoded the packed Day 19 and Day 20 overlays using their documented layouts. Day 19 uses ror8(byte XOR k2, 3) XOR k1 and checks the unpacked SHA-256. Day 20 checks the RS20 bounds and scratch SHA-256 before reading the HCFG resource.
  7. Generated every response from the extracted bytes. JSON-array answers were serialized as compact JSON strings because the submission API requires every answer value to be a string. Each submission used the exact IDs from its start response.

The implementation is in analyze.py. fetch_artifacts.py verifies downloaded delivery hashes; solve_mastery.py prepares the set-prefixed 4,704-field answer map. Dependencies and rerun commands are in README.md.

API results and artifact integrity

Single Solve

Artifact: Download evidence.zip SHA-256: 891cf4ebc9c0e8867641ad376c29f79a59a19781876d6ebad17b14c124e6ee5d

The server accepted all 392 answer fields. One useful cross-check was Day 1’s recover_code: the symbol’s bytes disassemble to mov eax, 0x1fa4; ret, so its unsigned decimal return value is 8100.

Time Trial

Artifact: Download evidence.zip SHA-256: 1e892688f9cfefe4f017bac816863cece7268c6c8cf127fdf5324d9d19032319 Deadline: 2026-10-28T13:29:32.131Z

All 392 fields were accepted before the deadline. The d10_analysis_section_count wording requested ELF e_shnum, but this delivery contained d10--sample.exe (challenge artifact, not published), a valid PE32+ x86-64 executable with nine COFF sections and no embedded ELF. Submitting the parsed PE section count 9 completed the attempt; the Game Master returned CORRECT.

Mastery

Mastery contained 12 separate evidence.zip (challenge artifact, not published) deliveries and 4,704 answer fields. All twelve downloads matched their API-provided SHA-256 values. The complete download links and hashes are in the Mastery brief; each delivery is also saved as mastery/delivery-NN.zip (challenge artifact, not published).

The final status reports all 4,704 answer IDs accepted and CORRECT, solved by Blandman3 at 2026-10-28T13:32:09.936Z. The submit response itself said the attempt was already solved; the follow-up status is the authoritative completion record.

Additional notes (merged from README.md)

Family Resemblance workbench

This directory contains the saved Game Master assignments, downloaded evidence, answer maps, analysis scripts, and the final write-up.

Reproduce the analysis

From this directory, install the parser dependencies:

python3 -m venv .venv
.venv/bin/pip install -r requirements.txt

analyze.py takes an AES-encrypted evidence.zip (challenge artifact, not published), extracts its members in central-directory order, validates embedded HCFG records, and writes ss_answers.json. The artifact ZIP uses the lab password infected.

For a saved Single Solve or Time Trial delivery:

FR_WORKDIR="$PWD" .venv/bin/python analyze.py

For the 12-delivery Mastery bundle, download and verify its assigned artifacts with the saved start response:

.venv/bin/python fetch_artifacts.py mastery/ss_start.json mastery \
  --challenge binary-reverse-engineering-day-28-family-clustering/mastery

Then derive the answer map:

.venv/bin/python solve_mastery.py mastery/ss_start.json mastery mastery/derived \
  --challenge binary-reverse-engineering-day-28-family-clustering/mastery

The default writes a reviewable ss_submit.json. Add --submit to send the answer map through the supported Game Master proxy. The saved mastery attempt is already solved, so a repeated submission will not score again.

render_assignments.py rebuilds the complete player-facing assignment briefs from the saved API start and status responses.

Files

  • REPORT.md — solution notes, findings, artifact hashes, and mode results.
  • mode-results.json — concise results confirmed by Game Master status.
  • CHALLENGE.md — full Single Solve questions, IDs, and formats.
  • Time Trial assignment — full timed assignment.
  • Mastery assignment — all 4,704 player-facing fields and 12 downloads.
  • analyze.py — extractor and answer generator.
  • fetch_artifacts.py — proxy downloader with SHA-256 verification.
  • solve_mastery.py — per-delivery solver and answer-map builder.
  • render_assignments.py — regenerates the full API briefs.
  • requirements.txt — Python dependencies.