Bitcoin Puzzle Solvers

Multi-Strategy Solvers for Bitcoin Challenge Transactions

---

Collection of solver implementations targeting the Bitcoin Puzzle Transaction challenge series. Includes algebraic, probabilistic, and heuristic approaches across CPU and GPU, with a CUDA kernel for batch scalar multiplication on RTX 4070 (Ada Lovelace, SM 89).

Repository Contents

bitcoin-puzzle-solvers/
├── bitcoin_sat_encoder.py           # Full SAT encoding of the privkey-to-hash160 pipeline (ECC + SHA-256 + RIPEMD-160)
├── pollard_rho_puzzle160.py         # Pollard Rho with partitioned walk function for Puzzle 160
├── puzzle135_hybrid_ai.py           # TensorFlow-guided hybrid solver for Puzzle 135
├── puzzle140_solver.py              # Targeted solver for Puzzle 140
├── puzzle160_16fold_attack.py       # 16th-root-of-unity symmetry reduction for Puzzle 160
├── puzzle160_crt_solver.py          # CRT-based decomposition solver
├── puzzle160_signature_search.py    # Signature-based pattern search
├── puzzle71_exhaustive.py           # Exhaustive enumeration for Puzzle 71
├── puzzle_solver.py                 # General-purpose puzzle solver framework
└── scanners/
    ├── adaptive_cpu.py              # Adaptive CPU scanner with dynamic range partitioning
    ├── ancient_node_finder.py       # Discovery of old Bitcoin nodes with weak validation
    ├── bitcoin_node_scanner.py      # Bitcoin P2P protocol node scanner
    ├── brain_wallet_recovery.py     # Brain wallet passphrase recovery
    ├── btc_personal_scanner.py      # Personal address monitoring scanner
    ├── btc_scanner.py               # General blockchain scanner
    ├── btc_scanner_v2.py            # Improved scanner with batch processing
    ├── epsilon_scanner.py           # Epsilon-neighborhood key scanner
    ├── epsilon_scanner_fast.py      # Optimized epsilon scanner
    ├── layered_scanner.py           # Multi-layer scanning strategy
    └── secp256k1_cuda.cu            # CUDA kernel: batch scalar multiplication (4x64-limb, modular arithmetic)

Technical Approaches

• Pollard Rho: Cycle-finding random walk on the elliptic curve group with configurable partition function
• SAT Encoding: Converts the full Bitcoin address derivation pipeline (scalar multiplication, SHA-256, RIPEMD-160) into conjunctive normal form
• Neural Hybrid: TensorFlow model trained to predict high-fitness regions for guided search
• 16-Fold Symmetry: Exploits primitive 16th roots of unity in the scalar field to reduce search space
• CUDA Kernel: 256-bit modular arithmetic with 4x64-limb representation, precomputed generator table, compiled for SM 89 (Ada Lovelace)

Requirements

• Python 3.10+
• Core: coincurve or ecdsa, gmpy2, numpy
• GPU: CUDA toolkit, nvcc with SM 89 support
• Optional: tensorflow, base58, psutil

Author

Andrew Dorman Independent Researcher -- Southlake, TX GitHub: ACD421