Neural Network Enhanced Blockchain Mining: A Modular-Exponentiation Proof-of-Work Framework with Learned Nonce Prediction

Read full text Summarise with VIRAT
Authors: Fiza Shaikh
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper presents Modular-Exponentiation Proof-of-Work (MEPoW), a formally defined consensus puzzle in which a valid block requires a nonce n such that SHA-256(n || header) satisfies a leading-zero target and the block-derived tuple (a, b, m, T) satisfies a^b mod m = T (mod m). Because (a, b, m, T) are deterministically derived from the block header, the modular residue condition is learnable. A dense neural network trained on (a, b, m) predicts a proximity nonce, replacing blind iteration with a targeted residual search. We formally prove the derivation bijection, quantify prediction-to-nonce proximity, and show that model leakage does not weaken MEPoW security beyond the inherent speedup bound. Experimental evaluation on 600 blocks across difficulty levels 3-8 shows mining-time reductions from 8.4x to 15.0x over single-threaded CPU brute-force, with energy per block of 11.4 J vs. 446.7 J (CPU baseline) at difficulty 6. With GPU inference the neural miner achieves 0.77 J/block -- 580x below CPU baseline and 17x below GPU brute-force. Mean prediction relative error is 2.61%, and a hybrid verification layer maintains a 100% valid-block rate across 600 test blocks. Ablation studies and a structured five-vector security analysis are included.

Fields of Study & Subjects

Blockchain Technology Applications and Security Cryptography and Data Security Cryptographic Implementations and Security

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
26 March 2026
Language
EN