| name | ctf-crypto |
| description | Solve cryptography challenges by fingerprinting the scheme, testing the weakest plausible attacks first, and scripting repeatable analysis. |
| license | MIT |
| compatibility | opencode |
CTF Crypto
What I do
- Classify encodings, ciphers, hashes, and mathematical structures.
- Distinguish between textbook crypto and challenge-specific mistakes.
- Test attack hypotheses from cheapest to most likely.
- Write solver scripts for transformations, brute force, and algebraic recovery.
When to use me
Use this for ciphertexts, keys, signatures, modular arithmetic, pseudorandomness challenges, and strange encoded blobs.
Workflow
- Establish what is encoding versus encryption versus hashing.
- Record all known values, lengths, alphabets, and formatting clues.
- Check for reused nonces, small exponents, weak randomness, oracle behavior, or structural leakage.
- Script experiments so results are reproducible.
- Confirm the recovered plaintext or flag format.
Parallel lanes
Crypto challenges often present multiple attack paths with varying cost. Run parallel hypothesis lanes:
Lane definitions
- Lane A: Encoding and simple transform checks (base64, hex, XOR with short/repeated key, ROT).
- Lane B: Statistical and frequency analysis for classical ciphers or weak randomness.
- Lane C: Mathematical structure tests (small RSA exponent, shared factors, linear congruential generators).
- Lane D: Oracle behavior probing if any service interaction is available (padding oracles, error leaks).
Lane budgets and search space limits
- Maximum 3 lanes concurrently.
- Lane A: 2-3 minutes (fast transforms only).
- Lane B: 5 minutes or until statistical anomaly found.
- Lane C: 10 minutes max; abandon if no algebraic structure emerges.
- Lane D: 20 queries max to oracle before requiring evidence of leak.
Merge criteria
- Partial plaintext recovered (even 10+ characters).
- Key material or private parameters recovered.
- Oracle behavior confirms leak (distinct responses to crafted input).
- Mathematical weakness confirmed (shared factor, small exponent, predictable RNG).
Kill criteria
- Search space exceeds feasible bounds (>2^30 operations without GPU/cluster).
- 50+ decryption attempts with no structure emerging.
- Oracle returns identical responses to 20+ distinct probes.
- Another lane has recovered partial plaintext.
Automation triggers
- Brute force search of key space (XOR keys, small RSA factors).
- Batch decryption of multiple ciphertexts.
- Oracle query automation (padding oracle, error oracle).
- Mathematical computation (GCD, factorization, lattice reduction).
High-value checks
- XOR with repeated or short keys.
- ECB or block-boundary leakage.
- RSA mistakes such as low exponent, bad padding, shared factors, or leaked structure.
- Linear congruential generator weaknesses.
- Custom schemes that are really simple transforms.
Rules
- Do not call something RSA, AES, or ECC without evidence.
- Measure before brute forcing.
- Track byte lengths and endian assumptions carefully.
- When math is involved, state the exact model and variables before solving.
