Pydome

Technical Walkthrough

Writeup

Challenge

• Name: Pydome
• Category: Misc
• Difficulty: Medium
• Mode: hybrid

Summary

Pydome is a Python input-filter challenge. The provided source hides the real ANACONDA environment seed, so direct seed recovery from the archive is not possible. The solve uses two service behaviors instead: SHA mismatch output leaks story.txt, and the non-integer conversion branch leaks deterministic random.randrange(100) outputs. A second leak plan with an initial pad-only attempt reveals the first hidden random value, allowing the first numeric-only secret and SHA target to be reconstructed.

Artifact Inventory

• files/a12c7371-a13a-4bc2-abcc-0565c45c4eb9.zip

- Contains only misc_pydome/server.py.

- The archive does not include story.txt, flag.txt, Docker metadata, or the ANACONDA environment value.

• Remote instance: <TARGET>:32587.
• Main evidence:

- analysis/extracted/misc_pydome/server.py

- analysis/remote/story.txt

- analysis/remote/random-leak-n100-repeat-indexes.json

- analysis/remote/prng-plan-20260610T235310Z.json

- analysis/remote/observed-prng-solve.json

Analysis

Source review showed four important facts:

1. The service requires exactly 100 comma-separated tokens.
2. Valid numeric tokens are used as indexes into story.txt; invalid numeric tokens can pad the 100-token length.
3. On SHA mismatch, the service prints the generated user_input, which turns story.txt into a remote oracle.
4. The suffix check uses zip(forest, user_input[0x40:]), so an empty suffix passes once the SHA target is correct.

The hidden SHA target is generated from Python random seeded with ','.join(os.getenv("ANACONDA").split("A")). The seed is not present in the provided archive, so broad seed guessing was closed after source/story guesses, system dictionary, and rockyou candidates did not match the live randrange(100) fingerprint.

The decisive path was to exploit observable random behavior. A leak100 attempt with 100 single-character non-integer tokens prints 100 chosen indexes from random.randrange(100). A separate plan beginning with pad,leak100,leak100,... produced a later printed window that aligned with the first leak at offset 50. The alignment equation identifies the first hidden random value as 31. For a numeric-only first solve attempt, that value is the secret length, and the first 31 observed leak values are the secret bytes. Hashing those bytes gives the required SHA target.

Solve

Reproducible scripts:

• solve/prng_oracle.py collects structured random leak windows.
• solve/solve_observed_prng.py infers the first hidden secret length from the leak alignment, builds the SHA target from observed PRNG values, encodes the SHA hex digest using indexes from the leaked story.txt, and submits the final numeric-only payload.

Final run:

cd <local workspace>
python3 scripts/challenge_exec.py Misc/Pydome -- python3 Misc/Pydome/solve/solve_observed_prng.py
python3 scripts/challenge_harness.py capture-flag Misc/Pydome --from loot/flag-candidate.txt

The solve transcript is saved at analysis/remote/observed-prng-solve.txt; raw flag material is stored only in loot/.

Flag

Raw flag is stored in loot/flag.txt and intentionally not reproduced here.

Lessons

• Do not keep guessing a hidden environment seed when the challenge exposes deterministic runtime behavior.
• A failed comparison that prints attacker-controlled reconstructed data can become a file oracle.
• Truncated PRNG leaks do not always require full MT19937 recovery. Here, one alignment between two controlled leak plans was enough to infer the first hidden length and reconstruct the needed first-attempt secret.

Source-Backed Dossier

The sections below are merged from companion Markdown notes for the same case. They are rendered after sanitization so the article stays precise without publishing raw flags, credentials, or target-specific secrets.

Notes

Scope

• Challenge: Pydome
• Category: Misc
• Difficulty: Medium
• Mode: hybrid
• Remote instance: <TARGET>:32587
• Start time: 2026-06-10T22:40:26Z
• Operator: harness
• State file: challenge-state.json

Harness Status

• Current phase: see challenge-state.json
• Next allowed actions: see next-action.json
• Raw flags and sensitive material stay in loot/ only. Do not paste them here.

Artifact Inventory

Evidence Ledger

Key Findings

• Extracted source is a single Python service: analysis/extracted/misc_pydome/server.py.
• The service requires exactly 100 comma-separated tokens.
• Valid numeric tokens index into remote story.txt; invalid large numeric tokens pad the token count without contributing output.
• On SHA mismatch, the service prints the generated user_input, which makes story.txt remotely leakable. The leaked story is saved at analysis/remote/story.txt.
• The leaked story is 1001 characters and contains all lowercase hexadecimal characters, so any SHA256 hex digest can be encoded as story indexes once the target digest is known.
• The forest check is weak: all(zip(forest, user_input[0x40:])) means an empty suffix passes if the SHA target is correct.
• The SHA target is generated from Python random seeded with ','.join(os.getenv("ANACONDA").split("A")); the ANACONDA value is not present in the provided files.
• Live PRNG fingerprinting works. analysis/remote/random-leak-n100.txt captures 100 randrange(100) outputs from the non-integer print path.
• Closed branches: obvious story/source seed guesses, system dictionary, rockyou, SHA256(empty) at attempts 1-5, and broad/noisy empty-state searches. See hypothesis-board.md and analysis/remote/*summary.json.
• Solved path: use solve/prng_oracle.py to collect observable PRNG windows, align the pad,leak100,... window against the first leak, infer the first hidden random value as the numeric-only secret length, hash the first observed leak values for that length, and submit the digest with solve/solve_observed_prng.py.
• Completion evidence: analysis/remote/observed-prng-solve.json shows flag_captured: true; the raw flag is stored only in loot/flag.txt.

RAG / Advisory Memory

RAG output is advisory only. Record evaluated retrievals with:

scripts/challenge_harness.py rag-record <workspace> --query "..." --tag MATCHED|PARTIAL|MISSING|<secret redacted>|GENERIC --validation "..."

Secrets/Flags

Raw flags and sensitive material stay in loot/ only. Use scripts/challenge_harness.py capture-flag to validate and record flag capture without printing the value.

Memory Summary

Metadata

• Platform: HackTheBox Challenges
• Category: Misc
• Challenge: Pydome
• Difficulty: Medium
• Source workspace: <local workspace>

Validated Solve Chain

Concepts only. Do not include raw flags, reusable credentials, tokens, cookies, private keys, or live secrets.

1.

Reusable Lessons

-

Dead Ends

-

Tool Quirks

-

Evidence Paths

-

Ingestion Decision

• Proposed for LightRAG: yes/no
• Requires user approval before ingestion: yes

Hypothesis Board

Keep no more than 3 active hypotheses on Easy/Medium and 5 on Hard unless the user explicitly asks for breadth.

Closed Branches

Memory Summary

approval_required: true

Sanitized Memory Summary

Metadata

• Platform: HackTheBox Challenges
• Category: Misc
• Challenge: Pydome
• Difficulty: Medium
• Source workspace: <local workspace>

Validated Solve Chain

Concepts only. Do not include raw flags, reusable credentials, tokens, cookies, private keys, or live secrets.

1.

Reusable Lessons

-

Dead Ends

-

Tool Quirks

-

Evidence Paths

-

Ingestion Decision

• Proposed for LightRAG: yes/no
• Requires user approval before ingestion: yes

Notes

Notes

Scope

• Challenge: Pydome
• Category: Misc
• Difficulty: Medium
• Mode: hybrid
• Remote instance: <TARGET>:32587
• Start time: 2026-06-10T22:40:26Z
• Operator: harness
• State file: challenge-state.json

Harness Status

• Current phase: see challenge-state.json
• Next allowed actions: see next-action.json
• Raw flags and sensitive material stay in loot/ only. Do not paste them here.

Artifact Inventory

Evidence Ledger

Key Findings

• Extracted source is a single Python service: analysis/extracted/misc_pydome/server.py.
• The service requires exactly 100 comma-separated tokens.
• Valid numeric tokens index into remote story.txt; invalid large numeric tokens pad the token count without contributing output.
• On SHA mismatch, the service prints the generated user_input, which makes story.txt remotely leakable. The leaked story is saved at analysis/remote/story.txt.
• The leaked story is 1001 characters and contains all lowercase hexadecimal characters, so any SHA256 hex digest can be encoded as story indexes once the target digest is known.
• The forest check is weak: all(zip(forest, user_input[0x40:])) means an empty suffix passes if the SHA target is correct.
• The SHA target is generated from Python random seeded with ','.join(os.getenv("ANACONDA").split("A")); the ANACONDA value is not present in the provided files.
• Live PRNG fingerprinting works. analysis/remote/random-leak-n100.txt captures 100 randrange(100) outputs from the non-integer print path.
• Closed branches: obvious story/source seed guesses, system dictionary, rockyou, SHA256(empty) at attempts 1-5, and broad/noisy empty-state searches. See hypothesis-board.md and analysis/remote/*summary.json.
• Solved path: <REDACTED>, align the pad,leak100,... window against the first leak, infer the first hidden random value as the numeric-only secret length, hash the first observed leak values for that length, and submit the digest with solve/solve_observed_prng.py.
• Completion evidence: <REDACTED>

RAG / Advisory Memory

RAG output is advisory only. Record evaluated retrievals with:

scripts/challenge_harness.py rag-record <workspace> --query "..." --tag MATCHED|PARTIAL|MISSING|<secret redacted>|GENERIC --validation "..."

Secrets/Flags

Raw flags and sensitive material stay in loot/ only. Use scripts/challenge_harness.py capture-flag to validate and record flag capture without printing the value.