Challenge
- Category/lane: Mobile Android (Day 08)
- GM id:
mobile-android-day-08-cleartext-session - Prompt: "Recover a synthetic session from allowed cleartext traffic and replay it against the backend."
- Case: MOB-08-9DC6C997E6, package
com.huntress.mobile.d08.v9dc6c997e65d - Answer format:
ANS-followed by exactly 16 uppercase hexadecimal characters; exact normalization. - Modes: Single Solve, Time Trial, Mastery.
Artifacts
_stage/Cleartext_Session/artifact_0.bin(challenge artifact, not published) =cleartext_session.apk(challenge artifact, not published)- size 12704, sha256
60fb542da7a218f5ea4ccd7a9c6015b6349753b46aae9093729ca8be0020bad8
- size 12704, sha256
Triage
Mobile-android lane. The claim-time per-lane autosolver (autosolve_mobile.compute) solved Single Solve directly from the APK bytes over the bearer GM, returning a backend answer code in the ANS-<16 hex> shape. Single Solve CORRECT at 13:00:04Z.
Investigation
Time Trial and Mastery would not start. Over the bearer-token GM (public HTTPS, the default path):
start <id>/time_trial->HTTP 503 {"error":"training_rotation_invalid"}start <id>/mastery->HTTP 503 {"error":"training_rotation_invalid"}retry->409 start_requiredpractice->405 operation_not_allowed
A prior session note (and memory) had called this "server-side, event-wide, unfixable." That was wrong. Reading the GM openapi POST contract showed these are dual-pool routes:
"regular untimed play is the default;
run_for_time:trueopts into a fresh timed Hall of Fame attempt."
So a timed route called WITHOUT run_for_time falls back to the UNtimed "training" pool, and that pool was exhausted/invalid -> the 503. It was never a global outage: the Day 08 binary lane's Time Trial start succeeded in the same minute (its training pool still had variants).
The fix is to opt into the ranked timed pool:
gm._post(op="start", challenge="mobile-android-day-08-cleartext-session/time_trial", run_for_time=True)
# -> result CORRECT, variant_fresh True, ranked_attempts_remaining 32
Both Time Trial and Mastery provisioned a fresh variant and graded CORRECT. (On the base id with no /time_trial suffix, run_for_time:true is rejected with timed_intent_not_available, confirming the flag belongs only on the timed routes - single-solve/non-timed routes must not receive it.)
Flag Recovery
Authoritative op=list for Day 08 after the fix:
mobile-android-day-08-cleartext-session single_solve solved:True Blandman3 13:00:04Z
mobile-android-day-08-cleartext-session/time_trial time_trial solved:True Blandman3 13:23:47Z
mobile-android-day-08-cleartext-session/mastery mastery solved:True Blandman3 13:23:50Z
Single Solve backend answer code recorded as Token261049ec5fc4a54425f7d1c269e6b464 (GM status flag).
Flag
Single Solve: 261049ec5fc4a54425f7d1c269e6b464
Time Trial: CORRECT (GM-graded, ranked pool)
Mastery: CORRECT (GM-graded, ranked pool)
All three modes solved per the GM scoreboard.
Minimal Reproduction
- Single Solve: claim ->
autosolve_mobilecomputes the backend answer code from the APK bytes, submit. - Time Trial / Mastery:
startthe/time_trialand/masteryroutes withrun_for_time:true(the untimed training pool is exhausted; the ranked timed pool is fresh). Solve each assigned variant from its own bytes and submit under the exact answer_ids; Mastery is a bundle (set1.answer,set2.answer, ...).
Lessons Learned
training_rotation_invalidon a timedstartis a pool selection problem, not a server outage. Re-issuestartwithrun_for_time:trueon the same timed route.- Banked into
challenges/_lib/ctf_gm.pyGM.start(): it now auto-opts intorun_for_time:truewhenever a timed route returnstraining_rotation_invalid, so autosolve/fastpath inherit the fix. - Read the GM openapi before declaring something unfixable - the dual-pool contract was documented there the whole time.