# US 2026/0189364 A1 · Tensor Network-Based Hash Function Attack

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0189364 A1
- Document ID: US20260189364A1
- Status: Published application
- Kind: A1
- Application number: 19/284,063
- Filed: 2025-07-29
- Published: 2026-07-02
- Earliest claimed: 2024-12-31
- Applicants: Multiverse Computing S.L.
- Inventors: Roman Orus; Borja Aizpurua; Sukhbinder Singh; Saeed Jahromi
- Web page: https://patentgrep.com/patent/US20260189364A1

## CPC

- H04L 9/0618: {Block ciphers, i.e. encrypting groups of characters of a plain text message using fixed encryption transformation}
- H04L 9/002: {Countermeasures against attacks on cryptographic mechanisms (network architectures or network communication protocols for protection against malicious traffic H04L63/1441)}
- H04L 9/0643: {Hash functions, e.g. MD5, SHA, HMAC or f9 MAC}
- H04L 2209/26: Testing cryptographic entity, e.g. testing integrity of encryption key or encryption algorithm
- H04L 2209/34: Encoding or coding, e.g. Huffman coding or error correction
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

A computer-implemented method for recovering a plaintext is disclosed. The plaintext (300) comprises a variable string (320). The method comprises constructing (S2) a tensor network (500) representing the variable string (320), optimizing (270) the tensor network (500), generating (S3), in a hash function generator (350), a candidate hash (370) based on the output of the tensor network (330), determining (S4) an overlap (390) between the candidate hash (370) and a reference ciphertext (380); and on reaching a zero-overlap value, determining the variable string (320), otherwise optimising the tensor network (500).

## Claims

### Claim 1

A computer-implemented method for recovering a plaintext, wherein the plaintext comprises a variable string and a fixed string, the method comprising:
constructing a tensor network representative of the variable string; and
iteratively optimizing the tensor network, wherein the iterative optimization comprises:
adjusting the internal parameters of the tensor network;
generating, in a hash function generator, a candidate hash from the fixed string and the output of the tensor network;
determining an overlap between the candidate hash and a reference ciphertext; and
when the overlap reaches a pre-determined threshold, determining the variable string and outputting the plaintext, otherwise further adjusting the parameters of the tensor network.

### Claim 2 (depends on 1)

The method of claim 1, wherein the variable string comprises a sequence of characters, and wherein the tensor network is created by mapping the characters of the variable string to a set of tensors and defining how the tensors connect to represent the relationship between the characters.

### Claim 3 (depends on 1)

The method of claim 1, wherein optimising of the tensor network comprises using a classical optimization algorithm.

### Claim 4 (depends on 1)

The method of claim 1, wherein the determining is carried out by calculating the Hamming distance between the candidate hash and the reference ciphertext.

### Claim 5 (depends on 1)

The method of claim 1, wherein the tensor network is one of a MPS tensor network or a flexible PEPS tensor network.

### Claim 6 (depends on 5)

The method of claim 5, wherein the tensor network is a MPS tensor network, the MPS tensor network comprising a series of tensors, wherein each tensor represents a segment of the variable string.

### Claim 7 (depends on 6)

The method of claim 6, comprising initializing the MPS network using a random distribution.

### Claim 8 (depends on 7)

The method of claim 7, wherein optimizing the tensor network comprises sweeping through the MPS tensor network from left to right and back, adjusting each tensor of the MPS tensor network sequentially, to iteratively recover the plaintext.

### Claim 9 (depends on 8)

The method of claim 8, wherein, for each tensor from the right-most to the left-most, the current tensor with its neighbour tensor is merged to form a combined matrix, and the combined matrix is updated by applying a random change, normalized, and then decomposed back into the original tensors via Single Value Decomposition SVD.

### Claim 10 (depends on 7)

The method of claim 7, wherein the internal parameters of the MPS tensor networks include hyperparameters including at least one of a virtual bond dimension of the MPS, a step length, a step, a reset value, a cutoff value, and temperature.

### Claim 11 (depends on 6)

The method of claim 6, comprising monitoring gradient norms are monitored to detect local minima; and if a local minima is detected, randomizing parameters of the tensor network, wherein the monitoring of gradient norms is done by monitoring an absolute value of the gradients.

### Claim 12 (depends on 1)

The method of claim 1, wherein the method is implemented by a classical processor.

### Claim 13

A system for recovering a plaintext, wherein the plaintext comprises a fixed string and a variable string, the system comprising:
at least one input/output device for inputting the plaintext and outputting the plaintext;
at least one tensor network, for encoding the variable string into a tensor network;
a hash function generator for creating a candidate hash from the fixed string and an output of the at least one tensor network;
a comparator for determining an overlap between the candidate hash with a reference ciphertext; and
at least one optimisation element for optimizing the parameters of the tensor network, wherein the at least one optimisation element is provided for sweeping through the at least one tensor network from left to right and back, adjusting each tensor of the tensor network sequentially, to iteratively recover the plaintext.

### Claim 14 (depends on 13)

The system of claim 13, wherein the tensor network is implemented as one of a MPS or a flexible-PEPS.

### Claim 15

A computer system for recovering a plaintext, wherein the plaintext comprises a variable string and a fixed string, the computer system comprising a classical processor configured to:
construct a tensor network representative of the variable string;
iteratively optimize the tensor network, wherein the iterative optimization comprises:
adjust the internal parameters of the tensor network,
generate, in a hash function generator, a candidate hash from the fixed string and the output of the tensor network;
determine an overlap between the candidate hash and a reference ciphertext; and
when the overlap reaches a pre-determined threshold, determine the variable string and outputting the plaintext, otherwise further adjust the parameters of the tensor network.

### Claim 16

A computer program product comprising logic for executing, by means of a classical processor, a method for recovering a plaintext, wherein the plaintext comprises a variable string and a fixed string, the method comprising the steps of:
constructing a tensor network representative of the variable string;
iteratively optimizing the tensor network, wherein the iterative optimization comprises:
adjusting the internal parameters of the tensor network,
generating, in a hash function generator, a candidate hash from the fixed string and the output of the tensor network;
determining an overlap between the candidate hash and a reference ciphertext; and
when the overlap reaches a pre-determined threshold, determining the variable string and outputting the plaintext, otherwise further adjusting the parameters of the tensor network.
