Tactical LogHuntress CTF 2026 // team archive

Days / Day 04 / Binary RE / Go Fish

DAY 04//LANE: BINARY RE//STATUS: CORRECT x3

Go Fish

Binary RE SSTTMM Modes reported CORRECT in this write-up

Challenge

Binary reverse engineering, Huntress CTF 2026 Day 04 (lane: binary-reverse-engineering, 32 fields). Prompt: "Go binaries are chatty: they ship with build info, function names, and file paths baked right in. Fish them out and map what this one was built to do." Modes: Single Solve (static), Time Trial, Mastery (2 evidence sets). Scored per field by the Game Master (partial credit, accepted fields accumulate).

Artifacts

artifact_0.bin (challenge artifact, not published) = AES zip (password infected, inert fixtures) holding four evidence items per set:

  • d01 dynamically linked x86-64 ELF (not stripped): header facts, section names, symbols, recover function disassembly.
  • d02 x86-64 ELF with an HCFG config blob.
  • d03 .NET PE32 DLL (+ runtimeconfig.json) with an XOR'd config.
  • d04 statically linked, stripped Go ELF (buildinfo present) with an HCFG config blob.

Triage

Same family as Day 03 "Table Manners" (dual-format RE inventory); new layer = the Go ELF. KB match: dual_format_re. Reused the banked solver, extended only d04.

Investigation

  1. Extracted the zip, hashed every file, ran file/readelf on d04 (ELF64, EXEC, little endian, EM_X86_64, Go build info at .go.buildinfo).
  2. Scanned d04 for the HCFG magic: layout is magic(4) + version(1) + XOR key(1) + len(2 LE)... same as Day 02/03; XOR the payload, CRC32 matches the header value. JSON holds campaign/domain/mutex/url.
  3. d01-d03 answers come from scripts/tm_base.py (copy of the Day 03 solver): section names, non-empty symbol names, mov eax, <recover_code> / ret, pe_cli / i386 / CLR metadata version for the .NET DLL.
  4. Submitted the 30 verified fields on the Single Solve (CORRECT). The two d04 label fields (architecture, format) were resolved in the Time Trial by submitting the sure batch first, then probing one field at a time: amd64 incorrect, x86-64 accepted; format elf accepted.
  5. Time Trial and Mastery ran as one automated shot each (scripts/run_gf.py), about 2 s end to end. Mastery needs per-set answers (setN. prefix); one 429 rate-limit hit mid-run was handled by adding submit backoff (start resumes the same attempt).

Flag Recovery

Each mode returned its completion token from the GM on CORRECT; values live on the GM/CourseStack and are not reproduced here.

Flag

Single Solve: CORRECT (completion token <REDACTED_TOKEN>)
Time Trial:   CORRECT (completion token <REDACTED_TOKEN>)
Mastery:      CORRECT (completion token <REDACTED_TOKEN>)

Minimal Reproduction

python3 "challenges/Day 04/Go Fish/scripts/run_gf.py" time_trial (and mastery) from the framework root inside WSL, with the GM token in place. Offline check: python3 .../solve_gf.py against extracted/.

Lessons Learned

  • Banked solvers cover the spine; only the new runtime (Go) needed work and the HCFG scan was format-agnostic.
  • Field labels for new fields: architecture = x86-64, format = elf. Probe label fields after a sure batch because the GM accumulates accepted fields.
  • Add 429 backoff to timed runners when several agents share one token.