Tensor network-based hash function attack

Tensor network techniques efficiently recover plaintext from hashed messages by iteratively optimizing unknown portions, addressing inefficiencies in existing methods and reducing computational requirements, suitable for classical hardware.

JP2026116806APending Publication Date: 2026-07-13MULTIVERSE COMPUTING SL
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2025254000
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-07-29
Filing Date
2025-12-16
Publication Date
2026-07-13

AI Technical Summary

Technical Problem

Existing methods for recovering plaintext from hashed messages are inefficient, especially for large messages or strong cryptographic algorithms, and require significant computational resources or quantum hardware.

Method used

A method using tensor networks, specifically Matrix Product States (MPS) and Flexible-PEPS Quantum Circuit Simulator (FQCS), iteratively modifies and optimizes portions of the plaintext to recover the complete message, leveraging known parts to focus the search, without requiring quantum hardware.

Benefits of technology

This approach significantly improves the efficiency of plaintext recovery by scaling better than brute-force methods, reducing computational resources, and enabling faster recovery of longer or more complex messages on classical hardware.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026116806000001_ABST
    Figure 2026116806000001_ABST
Patent Text Reader

Abstract

To provide a computer method for recovering plaintext. [Solution] The plaintext (300) includes a variable string (320). The method includes constructing a tensor network (500) representing the variable string (320) (S2), optimizing the tensor network (500) (270), generating a candidate hash (370) in a hash function generator (350) based on the output of the tensor network (330) (S3), determining the overlap (390) between the candidate hash (370) and the reference ciphertext (380) (S4), and determining if the variable string (320) is reached when a zero overlap value is reached, and optimizing the tensor network (500) if it is not.
Need to check novelty before this filing date? Find Prior Art