2026-10-02
Day 2: Mangled
Lane: Binary reverse engineering. Modes: Single Solve, Time Trial, Mastery. Result: all CORRECT.
The challenge
- ID:
binary-reverse-engineering-day-02-cpp-and-pe-orientation - Prompt: "C++ on Windows. The symbols are mangled, the classes hide behind vtables, and RTTI is your Rosetta Stone. Demangle it all with
c++filt, reconstruct the classes, and don't forget the PE imports."
What we were given
- Single Solve: one
evidence.zip(5,481 bytes, SHA-2567c4e11a97d97bda53d42ec4ea8ed096fec210454daaaf24f29329c5b9a418a13). - From Day 2 on, the zip files use AES with the password
infected. Extract with7z x -pinfected. - Two files:
day01-c.elf(a plain C program, like Day 1) andday02-cpp.elf(a C++ program). - Other modes: the C++ file was sometimes a Windows program instead (
day02-cpp.exe, PE32+). - Extracted files are saved in each mode folder.
What was asked
- 17 questions per delivery. Mastery had two deliveries, so 34.
- Every answer had an exact format. Use the literal
noneonly when a field is absent. - Question types:
- C file: SHA-256, the
.huntressmarker, whether.symtabexists, the constantrecover_codereturns, and the CPU architecture. - C++ file: SHA-256, container format (
ELForPE), and whether a vtable and typeinfo for the class exist. - C++ symbols: the mangled
scoremethod name and the vtable symbol name. - Config (new): the XOR key, the CRC32, and the
campaign,domain,urlandmutexfields of a hidden config.
How it was solved, start to finish
- Started the attempt (
7b599670-77a3-4f4e-8c5f-a611ec6a8bb2) and downloaded the zip. - Extracted it with
7z x -y -pinfected -oextracted evidence.zip, then hashed both files withsha256sum. - C file (same steps as Day 1):
readelf -p .huntressgave the markerBR01-CATALOG-000.objdump -M intel -d --disassemble=recover_codeshowedmov eax,0x1fa4, so the answer was 8100.
- C++ symbols:
nmlisted the mangled names.c++filtshowed what they mean:
_ZNK15DiscoveryEngine5scoreEPKcisDiscoveryEngine::score(char const*) const._ZTV15DiscoveryEngineisvtable for DiscoveryEngine._ZTV(vtable) and_ZTI(typeinfo) symbols exist, so both "has" answers weretrue.
- Config (the new part): the C++ file has a section named
.hcfg. It holds anHCFGrecord: - Bytes 0–3 are
HCFG. Byte 4 is the version. Byte 5 is the XOR key. - Bytes 6–7 are the payload length and bytes 8–11 are the CRC32 (both little-endian). The payload starts at byte 12.
- XOR each payload byte with the key, check the CRC32, and parse the result as JSON.
i = data.find(b'HCFG')
key = data[i+5]; plen, crc = struct.unpack_from('<HI', data, i+6)
payload = bytes(b ^ key for b in data[i+12:i+12+plen])
assert zlib.crc32(payload) == crc
print(json.loads(payload))
Re-checked on the saved file with hcfg.py:
=== HCFG valid records in day02-cpp.elf ===
offset=0x2180 version=1 xor_key=0x31 plen=367 crc32=0x6149aa00
JSON: {"campaign":"BR02-000","domain":"node-000.day02.example.invalid","ipv4":"198.51.100.40","url":"https://node-000.day02.example.invalid/api/BR02-000","mutex":"Global\\HuntressBR02_000", ...}
- Submitted the 17 answers in
answers.json. Examples:"cpp_xor_key": "0x31","cpp_config_crc32": "0x6149aa00","vtable_symbol": "_ZTV15DiscoveryEngine". Result:CORRECT. - Time Trial:
- The first attempt (
d51ac797-…) ran out of time (statusexpired). - Retry 2 shipped the C++ program as a PE file. Its submit came back
incorrect. - Retry 3 (
a01b1fde-9885-4ccf-8cc2-b7f64dbc9adc, zip SHA-25671b55cec3c530a3dd025701cf3fe1c43b6afa8fc83eb647e49629ef67e8dde0a) shipped an ELF again. Its config had XOR key0x32, CRC320xffdeebbeand campaignBR02-001. Result:CORRECT.
- Mastery:
- Attempt
7cd50eae-6904-4df0-a662-56bac1ac35d8gave two deliveries: - Delivery 1 (SHA-256
8c78c4eaa6235a25e4b3f033b6e5d562c0d321889fd0a4459be723158479d1ca):day01-c.elfandday02-cpp.exe(PE32+). - Delivery 2 (SHA-256
9b07769d1bc97535a8a1294ba506ca666845937ae2cc1fb4d232ba1bc9d37815):day01-c.elfandday02-cpp.elf. - For the PE file, the symbol answers are the COFF section symbols shown by
objdump -t:.text$_ZNK15DiscoveryEngine5scoreEPKcand.rdata$_ZTV15DiscoveryEngine. Its config: XOR key0x32, CRC320xffdeebbe, campaignBR02-001. - Delivery 2: marker
BR01-CATALOG-005,recover_code8105, XOR key0x36, CRC320xe810e785, campaignBR02-005. - The timed attempt expired. The Game Master allowed untimed practice (
op=practice) on the same files. All 34 answers wereCORRECTthere.
Results
| Mode | Result | Flag |
|---|---|---|
| Single Solve | CORRECT | 95c8641ef1f5afd697437e2976468986 |
| Time Trial | CORRECT | 761b7c0feecc075175e39573ea0c9a1c |
| Mastery | CORRECT (practice) | c2520437aa1b82cde610d9fda7723627 |
Each mode folder has the zips, the extracted files, the answers, commands-output.md, and the raw Game Master responses. The Time Trial retries are saved as retry-2-* and retry-3-*. No screenshots were taken.
Note: single-solve/commands-output.md prints the zip hash with two characters missing. The correct value is the one above (it matches start-response.json and sha256sum).
Reproduce
cd single-solve
7z x -y -pinfected -oextracted evidence.zip
sha256sum extracted/*
readelf -p .huntress extracted/day01-c.elf
objdump -M intel -d --disassemble=recover_code extracted/day01-c.elf
nm extracted/day02-cpp.elf | grep -E 'score|vtable|typeinfo'
c++filt _ZNK15DiscoveryEngine5scoreEPKc _ZTV15DiscoveryEngine
python3 -I ~/.config/huntress/solvers/hcfg.py extracted/day02-cpp.elf
These commands only read the files. They never run them.
Day 2: Looks Can Be Deceiving
Lane: Metadata forensics. Modes: Single Solve, Time Trial, Mastery. Result: all CORRECT.
The challenge
- ID:
metadata-forensics-day-02-looks-can-be-deceiving - Prompt: "The extensions claim one story; the first sixteen bytes tell another. Identify what each artifact actually is and isolate the deliberate disguise."
- Attempts:
- Single Solve:
6b5c078c-9c8a-452d-9f5a-ebbd4a2bdcd4, started2026-10-02T16:24:45.270Z, untimed. - Time Trial:
66ea1448-7be0-4e0f-951f-43393157d222, deadline2026-10-02T16:41:10.321Z. - Mastery:
6ea31609-14ce-4f0c-b7e1-62526d102757, deadline2026-10-02T16:43:18.445Z.
What we were given
- Single Solve:
(287 bytes, SHA-256metadata-day-02-evidence.zipc356020d0dafda86d80d000b3e3e1c0524f3a0c1a84ade856a74959047c1be67). One member:(126 bytes, SHA-256evidence-note.txt672840e4d6e79fdb346d91735916cdc6ab52abd2aa0371bc6d5eb20f65aaedd8). - Time Trial:
(SHA-256metadata-day-02-evidence.zipcb06ac9e8c8a73195752578d821dd7ce8cb83d42cf1fb123bc05e58af2a8cdfb). One member:(SHA-256roster.csvc9f99583d9800af8ff9945904dbe0626046464f1510715db10effd3c32f1a5ee). - Mastery:
(SHA-256mastery-bundle.zip286fe65f5037df6f4b42c2404d56e50a6156ae03635349f4f81b39a474320c98). It holds two inner zips: (SHA-256metadata-day-02-evidence-01.zip8bf333163256a4a14078dcc81c31ca56fd9e73ff263452cd494f74a01a56c675) with(SHA-256evidence-note-copy.txt033c3aa4520d957dbf97964411082f78ce37ed957106d88e2ae5ca774c8864cd).(SHA-256metadata-day-02-evidence-02.zip96434ff85c149297d040b4890a8513c486d226736d92380759c01cb16a97f03f) with(SHA-256disguised-report.binbb62fbe0db033ae5cccc88054d3dde09d995223d5d9680e5fb37c4701df95cdf).
What was asked
- One question type: "What are the first sixteen bytes as lowercase hexadecimal?"
- Format: exactly 32 hex characters, no
0xprefix, no spaces. The answer is trimmed. - Answer IDs:
- Single Solve:
evidence-note-magic-prefix-hex-16. - Time Trial:
roster-magic-prefix-hex-16. - Mastery:
set1.evidence-note-copy-magic-prefix-hex-16andset2.disguised-report-magic-prefix-hex-16.
How it was solved, start to finish
- Started the attempt and noted the answer ID from the reply.
- Downloaded the zip and extracted its member. For Mastery, opened the outer bundle first, then each inner zip.
- Read the first 16 bytes of each assigned file and printed them as hex:
python3 -c 'from pathlib import Path; print(Path("single-solve/files/evidence-note.txt").read_bytes()[:16].hex())'
- Results per file:
| File | First 16 bytes (hex) | As text |
|---|---|---|
evidence-note.txt | 48756e7472657373204d657461646174 | Huntress Metadat |
roster.csv | 616e616c7973745f69642c636173655f | analyst_id,case_ |
evidence-note-copy.txt | 48756e7472657373204d657461646174 | Huntress Metadat |
disguised-report.bin | 255044462d312e370a252048756e7472 | %PDF-1.7\n% Huntr |
- Found the disguise.
disguised-report.binclaims to be a.binfile, but it starts with%PDF-1.7. It is a PDF. - Submitted one JSON object per mode. Mastery:
{
"challenge": "metadata-forensics-day-02-looks-can-be-deceiving/mastery",
"answers": {
"set1.evidence-note-copy-magic-prefix-hex-16": "48756e7472657373204d657461646174",
"set2.disguised-report-magic-prefix-hex-16": "255044462d312e370a252048756e7472"
}
}
- The Game Master replied
CORRECTeach time and returned the flag.
Results
| Mode | Result | Flag |
|---|---|---|
| Single Solve | CORRECT | 3438d22184f1243a9c8fc30cdc9f54e9 |
| Time Trial | CORRECT | ce928a0b41a1376a68f70be80133f63b |
| Mastery | CORRECT | d6004ae265e59a49f8cf1efd57d4bc22 |
This folder keeps the zips and extracted files only. The Game Master JSON replies for Day 2 were not saved.
Reproduce
python3 -c 'from pathlib import Path; print(Path("time-trial/files/roster.csv").read_bytes()[:16].hex())'
python3 -c 'from pathlib import Path; print(Path("mastery/files/metadata-day-02-evidence-01.zip.contents/evidence-note-copy.txt").read_bytes()[:16].hex())'
python3 -c 'from pathlib import Path; print(Path("mastery/files/metadata-day-02-evidence-02.zip.contents/disguised-report.bin").read_bytes()[:16].hex())'
These only read bytes. run_meta.py later handles the same field type (-magic-prefix-hex-16 returns fb[:16].hex()).
Day 2: Deep Link Drift
Lane: Mobile (Android). Modes: Single Solve, Time Trial, Mastery. Result: all CORRECT.
The challenge
- ID:
mobile-android-day-02-deep-link-drift - Prompt: "Day 1 taught you the roads. Now learn which ones are actually guarded — tell verified routing apart from a scheme and host that wave everybody through."
What we were given
- One Android APK per case, 12,704 bytes each. Inside:
AndroidManifest.xml,classes.dex,assets/huntress_puzzle.json, andMETA-INF/signing files. - Single Solve:
deep_link_drift.apk, caseMOB-02-15B6E1B835, packagecom.huntress.mobile.d02.v15b6e1b8356f, SHA-2569211e52f1844b86ef8f75622dc36df52e6e7e9935b48cf1d24ded475148cac07. - Time Trial: saved as
time-trial/deep_link_drift.apk, caseMOB-02-7961FE2922, packagecom.huntress.mobile.d02.v7961fe2922d9, SHA-256cb6af40479cce787a74b3c4394469678383935ec1b99e1deeb5ca6e7a42d91de. - Mastery: three APKs in
mastery/delivery-1todelivery-3: deep_link_drift_mastery_01.apk, SHA-256e5c2b3a7ebf4473459b5f77720dd1037bfee87fbd85a4524e471a9a5fbf8183bdeep_link_drift_mastery_02.apk, SHA-25688fae57a5468a98fdf431027b17a45d68bfb9c8615bf55dc2468a9e0a4eac658deep_link_drift_mastery_03.apk, SHA-256398e080dd91f4d2b8fc12147fb5c686aed9b08b0261f3f6875a6e1562fb58198
What was asked
- One question per case: return the exact backend answer code for that case.
- Answer format:
ANS-plus exactly 16 uppercase hex characters. - Answer IDs:
answerfor Single Solve and Time Trial.set1.answertoset3.answerfor Mastery. - Time limits: Time Trial 50 seconds (deadline
2026-10-20T14:02:07.789Z), Mastery 110 seconds (deadline2026-10-20T14:03:29.591Z).
How it was solved, start to finish
The app was never installed or run.
- Started the attempt (Single Solve attempt
ef3f32cd-7b0e-495d-859b-289cc82175d3) and downloaded the APK. - Checked the hash with
sha256sum. It matched the start response. - Read the manifest with
aapt dump xmltree. It shows an exported.ChallengeActivitywith schemehuntress02, hostdeep-link-driftand path prefix/unlock. - Read the puzzle asset:
unzip -p deep_link_drift.apk assets/huntress_puzzle.json
{"schema":"huntress.mobile.puzzle.v1","day":2,"codec":"reverse","data":"94AEE0A765B0097A-SNA"}
- Decoded it. New codec today:
reverse. Reverse the string:94AEE0A765B0097A-SNA→ANS-A7900B567A0EEA49. - Submitted under
answer. The Game Master repliedCORRECT. - Time Trial (attempt
0fd9e1f6-0216-45c2-b708-0a316040daf3): same steps.44466DC53283AEC9-SNA→ANS-9CEA38235CD66444. - Mastery (attempt
9fed887b-e073-45ef-8b9b-d48d2edfaadc): reversed the asset in each APK, then sent all three in one submit.
set1.answer MOB-02-285188CF6B data 28E64C9FF60E34F9-SNA answer ANS-9F43E06FF9C46E82
set2.answer MOB-02-5144692789 data 519F497B87D0CC07-SNA answer ANS-70CC0D78B794F915
set3.answer MOB-02-47421FF0E6 data 4E80D7F727DE7C25-SNA answer ANS-52C7ED727F7D08E4
Results
| Mode | Result | Flag |
|---|---|---|
| Single Solve | CORRECT | 7d214fa8d5d1785a754c73374679bccb |
| Time Trial | CORRECT | b4e4e338ff1748f180d36857d2550db6 |
| Mastery | CORRECT | 64d3d8f9f9e8494dbacaeb4203c7f08a |
Each mode folder has the APK(s), answers.json, start-response.json, submit-response.json, final-status-response.json and a commands-output.md. No screenshots were taken.
Reproduce
Offline, from the saved APK:
unzip -p single-solve/deep_link_drift.apk assets/huntress_puzzle.json
python3 -I -c 'print("94AEE0A765B0097A-SNA"[::-1])'
A fresh run with the later lane driver (it handles the reverse codec):
python3 -I ~/.config/huntress/solvers/run_mode.py mobile-android-day-02-deep-link-drift single_solve mobile_puzzle /tmp/d02
This needs a Game Master token in ~/.config/huntress/gm-token. It only unzips the APK and reads the asset. It never installs or runs the app.
Day 2: Host With the Most
Lane: Script analysis. Modes: Single Solve, Time Trial, Mastery. Result: all CORRECT.
The challenge
- ID:
script-analysis-day-02-encoded-intent - Prompt: "Windows Script Host has JScript plans for execution, download, upload, and impact. Carry your PowerShell findings forward, tell a function call from an invocation or a property access, and work out which script hands off to which."
What we were given
- One
evidence.zip(SHA-2567743a1af7c51d0950e7620561cf392791a9537e044a8ca11d344b11540ade516). Password:infected. - 12 files:
- The same three PowerShell samples as Day 1 (
resource-,collection-,impact-sample.ps1). - Nine JScript files. The file name tells you which role to look for:
| File | Role it plays |
|---|---|
host-context.js | benign |
memory-stage.js | stager |
embedded-write.js | dropper |
remote-write.js | downloader |
local-dispatch.js | launcher |
outbound-transfer.js | uploader |
impact-plan.js | impact |
staged-handoff.js | chain |
relay-handoff.js | relay |
- Each JScript file packs a config with its own trick: a Base64 string with a hand-written
decodeBase64, number lists split on:or-,^ keyXOR loops,.reverse(),keyBytesorstoredMaterialarrays. Unpacked, every file gives the same shape:{steps, adapter, values, stageCount}.
What was asked
- 87 questions: 15 for the PowerShell files (same as Day 1) and 72 for the JScript files (8 per file).
- Per JScript file:
- Role: classify by the ordered behavior graph (
benign_utility,stager,dropper,downloader,uploader,launcher,impact,staged_chain, ...). - ATT&CK ID, or
nonewhere allowed. - Adapter, campaign ID, and one role-specific indicator (URL, path, next-stage file, extension or note name).
- Three syntax questions: function style (
declaration,anonymous_expression,named_expression), invocation style (direct,call,apply) and property style (dot,bracket_literal,mixed). - Small format notes from the Game Master:
- Answers are trimmed before they are checked.
- For
jscript-benign-rolethe question text lists 8 roles, but the accepted values also includerelay,persistence_descriptorandadvanced_syntax.
How it was solved, start to finish
- Started the untimed attempt
dd3825ea-49e3-4c47-a202-60d367f89adaat2026-10-02T16:46:32Z. - Downloaded and extracted
evidence.zipwith 7z. The extracted files are saved insingle-solve/. - PowerShell: decoded the Base64 JSON descriptor in each
.ps1file, the same way as Day 1. - JScript: decoded each config statically, then read three things from it:
- the ordered
steps(the behavior graph), - the
adapter, - the
values(campaign ID and indicators).
The scripts were never run. The decoded graphs and the saved answers (re-checked offline with tools/jsdec_py_emu.py):
| File | Decoded steps | Adapter | Role | ATT&CK |
|---|---|---|---|---|
host-context.js | decode_configuration>return_configuration | Scripting.Dictionary | benign_utility | none |
memory-stage.js | acquire_remote>dispatch_memory | MSXML2.XMLHTTP.6.0 | stager | T1105 |
embedded-write.js | decode_embedded>write_file | ADODB.Stream | dropper | T1140 |
remote-write.js | acquire_remote>write_file | MSXML2.XMLHTTP.6.0 | downloader | T1105 |
local-dispatch.js | locate_local>dispatch_process | regsvr32.exe | launcher | T1059.007 |
outbound-transfer.js | locate_local>prepare_transfer>upload_remote | CertReq.exe | uploader | T1041 |
impact-plan.js | enumerate_targets>transform_content>write_recovery_note | ADODB.Stream | impact | T1486 |
staged-handoff.js | acquire_remote>write_file>dispatch_process>handoff_script | Microsoft.XMLHTTP | staged_chain | T1059.007 |
relay-handoff.js | same as above | WinHttp.WinHttpRequest.5.1 | staged_chain | T1059.007 |
- Handoffs ("which script hands off to which"): the chain and relay files name their next stage.
staged-handoff.jshands off toresource-sample.ps1(another language).relay-handoff.jshands off tolocal-dispatch.js(same language). - Syntax questions: read straight from the source text. Examples:
function inspect_x(...)is a declaration, andvar inspect_x = function(...)is an anonymous expression. A.call(or.apply(on the descriptor-recovery call sets the invocation style. - Answers. A few real ones from
single-solve/answers.json:
{"jscript-stager-role": "stager", "jscript-stager-attack": "T1105",
"jscript-stager-adapter": "MSXML2.XMLHTTP.6.0", "jscript-stager-campaign": "campaign-a3a04f5ced",
"jscript-stager-url": "https://a3a04f5ced.example.invalid/36925272.dat",
"jscript-stager-function-style": "anonymous_expression", "jscript-stager-invocation-style": "apply",
"jscript-stager-property-style": "dot",
"jscript-chain-next-stage": "resource-sample.ps1", "jscript-relay-next-stage": "local-dispatch.js"}
- Submitted all 87 answers under their exact IDs. The Game Master replied
CORRECT. - Time Trial: attempt
c38b2d1e-db06-44d6-ad00-ff6126e377b5. The deadline was 5 seconds after the start (2026-10-02T17:05:59.336Z). Newevidence.zip, SHA-2569bb6e0a84469b3f4695cc3acf6f99dccfb8e3f4a6fa650357a1d6871592b3423. Same steps. The client measured 1.48 seconds from retry to submit. - Mastery: attempt
d930eed5-fdbd-46ec-b996-945353925bb0, 174 questions in two sets. The deadline was 6 seconds after the start (2026-10-02T17:11:10.515Z). Two archives:
- set 1: SHA-256
0800fdb496f3bba7cf3f25df107c13ba12d00c093b95757fd0edd5fab40dbb27 - set 2: SHA-256
19f4279da8ebfd42fa9c6a9ebb8aa71e67e96e28326497c34b15be0414f83d31
Both sets were decoded and submitted together. The client measured 1.84 seconds.
Results
| Mode | Result | Flag |
|---|---|---|
| Single Solve | CORRECT | 3b388ccd7af0941e48b66023d9721370 |
| Time Trial | CORRECT | f02d8773484e5485c18a507164564777 |
| Mastery | CORRECT | 4a858c1eee015ee32e66489eda62cef8 |
Saved responses:
- Single Solve:
challenge-details.json,start-response.json,status-response.json. - Time Trial:
retry-response.json,status-response.json,submit-response.json. - Mastery:
attempt-response.json,status-response.json,submit-response.json.
Each mode also keeps its archives and extracted files. No screenshots were taken.
Reproduce
Offline, from the repo root, using the saved Single Solve archive:
mkdir -p /tmp/d02/ex
7z x -pinfected -o/tmp/d02/ex repos/day-02-script-analysis-host-with-the-most/single-solve/evidence.zip
python3 -I tools/solve_expansion.py /tmp/d02/ex \
repos/day-02-script-analysis-host-with-the-most/single-solve/start-response.json /tmp/d02/answers.json
This rebuilds all 87 saved answers (re-checked on 2026-11-01). It reads the scripts as text only. In the tools/ copy, JScript is decoded by jsdec_py_emu.py (pure Python).
Day 2 — Half a Turn (ROT13)
Lane: Warmups · Challenge ID: warmups-day-02
> Thirteen steps forward. Funny how that works out.
What we were given
A text blob to decode. Each mode provides one or more blobs.
How it was solved
The title "Half a Turn" and the hint "thirteen steps" point to ROT13 — rotating each letter 13 positions in the alphabet (half of 26).
- Downloaded
blob.txtfrom the Game Master. - Applied ROT13 decoding to the entire file contents.
- The decoded text ends with a slug in the format
word-word-word-NN. - Submitted the slug as the answer.
For mastery, 8 blobs were provided in a zip file. Each was independently ROT13-decoded and its slug submitted.
Automated with warmup_decode.py — a general-purpose decoder that handles ROT13, hex, base64, and 25+ other encoding types across the warmup lane.
Results
| Mode | Result | Flag |
|---|---|---|
| Single Solve | CORRECT | aee569db2b23f0dfcffd232b700eac37 |
| Time Trial | CORRECT | 7f00336a7a5cd16995653ba5544878fe |
| Mastery (8 blobs) | CORRECT | 4a88ed9065ca958928bfb5612f591dfd |