Tactical LogHuntress CTF 2026 // team archive

Days / Day 23 / Mobile Android / Hook Line and Signal

DAY 23//LANE: MOBILE ANDROID//STATUS: CORRECT x3

Hook Line and Signal

Mobile Android SSTTMM Modes reported CORRECT in this write-up

Day 23 - Hook Line and Signal (mobile-android)

Agent: agent-5. All three modes CORRECT (Blandman3).

ModeResultFlag
Single SolveCORRECT97ac85d15320be6d762dce8a9130e5b3
Time TrialCORRECTe202b99fb448a39b44855f2a3f09f289
MasteryCORRECT (ranked)a2e67d6bbb7c09e0855a667658853f18

Single

assets/huntress_puzzle.json used the known xor codec; the banked autosolve_mobile decoded it in 1.8 s (AnswerANS-75C47B4BF8B5B1C1).

Time Trial / Mastery - new layer: codec analysis

The puzzle JSON carried no data at all (schema huntress.mobile.puzzle.v2, day, codec: analysis, build). The answer is not stored anywhere; it is computed by the APK's own code: Release.code() = Release.token() -> Mix.f1 -> Util.fin* -> xor constant -> 16 uppercase hex nibbles prefixed ANS-. token() is a chain of static long mixers (Util.mix*, Util.scan*/Core.scan*/Util.walk* using Probe.s2/s3 loops and Tab.build lookup tables, HookScan/Guard.merge*). Everything is pure 64/32-bit arithmetic with no runtime inputs (the proof / client_secret / 127.0.0.1:19090 evidence URL are a decoy path; the Game Master question only wants the ANS code).

Solution (static only, nothing from the sample is executed):

  1. baksmali disassembles classes.dex (challenge artifact, not published).
  2. challenges/_lib/smali_static.py interprets the static methods of the dump (const, move, int/long ALU ops, shifts, if/goto loops, static calls, int/long arrays via new-array/aput/aget, return-object) and evaluates Release.code() up to the StringBuilder; v0 is formatted as ANS-%016X.
  3. Wired into autosolve_mobile._analysis() (also dumps each APK to challenges/Day 23/scripts/analysis_<ts>.apk).

Mastery bundled several APKs from other days' themes (Day 2, Day 12, ...) all using the same analysis codec; the evaluator handled them after adding array support (new-array/aput/aget) and return-object.

Timeline / mistakes

  • First timed attempt opened before the extractor existed (codec unknown), burning one Time Trial window; an untimed continuation then completed the Time Trial. Mastery's first ranked window was lost to an unsupported opcode (new-array) and later one (return-object); after both were added the next Mastery run was accepted by the GM as CORRECT (flagged RANKED).
  • Lesson: when a mobile puzzle has codec: analysis and no data, disassemble Release.code() and evaluate it statically.